Feedback: I have tried to do a lot of things through automation that I eventually realized couldn't fully be done in an automated way. Or, put another way, I could get like 80% of the way there with automation and the remaining 20% had to be done by hand. Which led me into sort of an 80/20 "rule" paradox: Getting 80% done only took 20% of the time. Getting the remaining 20% took 80% of the time. After making a scratch area (looks like this) and turning the shadows on and off in the toolset I was seeing very few shadows. But what really stood out were the smoothing discrepancies. I understand that given the overall nature of the material itself, there are going to be "smoothing discrepancies" because it's granite and it cleaves in such and such a way and it's just going to look like that. But the discrepancy between tiles (if you look on the right side of that pic) is really pronounced. To take a sample tile: tmd07_0011_2. It looks kind of wonky on the slope and the tile seams are really visible. The part of the tile near the top of the picture looks freaking fantastic but the bottom part and the edge seams maybe deviate away from looking like granite and more just like a low poly model with not a lot of smoothing applied. From reading upthread it looks like you might be doing an autosmooth set to 10 (?) over the whole tile.
I'd just kick out two pieces of feedback which I admit are of dubious utility because I'm no tile expert:
First, would it be possible to apply a higher smoothing to faces closer to the edge and less in the middle 3/4 (or 7/8?) of the tile to make the issue I describe above "go away" while still retaining the automated nature of the process?
Second, I kept ticking shadows on and off and saw very few shadows appearing- it could obviously have been my choice of tiles. However, there was a self shadowing (I think that's what it was) on one tile and another which was close to a body of water was throwing a weird shadow fragment on it. If you had 10 "zots" to throw at the tileset between shading and shadows, my vote would be more zots in the smoothing and getting the tiles to look "just so" and not spend so many zots on the shadows, which are (I'm assuming) way more time consuming per fix. You had sort of floated a question earlier about how many of us used shadows. For me, if something comes with shadows I'll probably play around with them on. But if they aren't included, no big deal. I just don't miss them that much. But what really stands out is the smoothing, because that's always on unless the ambient sun/moon is set so light that it blows it away.
Keep in mind this is still WIP. I'm sure you did, and thanks for the feedback.
In comment to the shadows on/off thing. I noticed some of the times that shadows from previous models changed to another will linger in the toolset. I expect this is some kind of error in the toolset not being reported. It happens far less now than before, but still does happen. If I start a chart of which tiles do this, I can probably find a common issue and fix it.
In comment to the top-down picture you sent with the lack of smoothing, I specifically am using a very slow smoothing threshold because using a larger one, even as much as 15, causes the tile, which is smoothed on a curve system, to toss that curve UP on one tile, and DOWN on the adjacent one. The only way to fix that is to manually modify the edge normals, which doesn't work correctly from export of GMax to import in NWN, or to make an amount of faces at the edge of any tile which duplicate the position and angle of all potential adjacent tiles. I have a script system which does this, but it doesn't work as well as you might think. Because smoothing is based on a curve, sometimes a single face is not enough to get the wave to come near a zero point. My script went so far as to try three faces at different lengths from the true edge of the tile. None of the combinations I came up with were sufficient, as they either required a distance I was uncomfortable sacrificing from the edge of the tile, or they actually made smoothing worse in on way or another.
Another thing I tried, which actually worked well, even though I got lots of suggestion NOT to do it, was to let the edge mesh flow outside the tile boundary box. This let me flow the contour from one tile into the next. There was still a seam, but that let the seam be something other than straight lines. Check out very early in this thread for screen shots in the more bubbly black hills tileset. I really tried to distance myself from that method with this script, but I still have the other scripts to make those paper-doll type wrapping meshes. One big thing I think that kind of mesh will do is destroy the whole shadow mechanism. Because up to half of the tile mesh is actually outside the tile boundaries, I can't even imagine one way or another if this shadow doctor script will work on such a monster.
Back to the pictures you linked. The most prominent tile edge mess is visible on majority of '_2 tiles. This is because my alt tile builder raises the center vert, all faces adjacent to that, and all faces adjacent to those. That actually spans to the verts second from the tile edge on the primitive mesh I was using. The problem is then caused when the outermost group of faces is not flush with the next tile. In the case of '_2 tiles, this is much more than simply not flush. It's more like 15-30 degrees off flush, so it is quite visible. I don't much like it either, but I need to rewrite that part of the script, possibly using less faces around the center part that I raise, or somehow splitting the outermost selection faces in half before I push them up. That would leave the half of those faces nearest the edge much more flush with adjacent tiles.
Just to reiterate, increasing the smoothing threshold around the edge will actually make it worse, not better, unless the ring around the tile is divided into a detached element first, removing the wave-like smoothing at the edge of the rising area.
On shadows: if you can point me to any one tile that throws spikes, I would gladly fix those, especially if they throw negative shadow spikes outside their tile box. But I want to make sure everybody knows the long white parts on the tiles right now are not negative shadow, but instead caused by the low quality shadows I am currently creating using the walkmesh. Because the walkmesh shape sometimes goes over the visible mesh, what you get is self shading. Most of the time the walkmesh surface falls under the visible mesh, and that is where shadows are the best. When I export this using shadows created from the visible mesh, the shadows are much better, 4x as complex, and it takes a long time to export. Since I'm not really pushing that work right now, I'm repeatedly exporting with the walkmesh-grade shadows. If you'd specifically like to take a look at the higher grade shadows, I can do one up tonight and pack it so you can see the quality and bloat difference. It's quite interesting. You will however see some places where the shadow rendering stuff in the engine totally fails. Sometimes you will see a shadow mesh face which is in the same position as a visible face. Since I have no way to decide where shadow faces will be drawn when the lighting is applied, I don't know how to fix that. Another issue is that sometimes the shadow faces are detached from the rest of the shadow, and float up to 5 cm above the shape they are supposed to be parallel to. This looks really weird. Between these two really noticeable issues, I would say the last time I exported full shadows, there was one face on every 30 tiles which had a flicker-face shared position, and there were about 1/10 tiles which had misplaced shadow faces. Again, I don't know how to fix that, since I don't have access to the code which actually draws the shadows.
So back to smoothing. One of the biggest ways I intend to overdub the visible geometry lines is to paint decals onto sections which I pop off with another script I am working on. This script will detect verts which are higher than the faces it is attached to. It will then spread out its selection a number of times, or less if the shape of the mesh begins to curve back upward. This lets me select and detach circular and eliptical regions, clone it, place it over itself by a single cm, and then paint it with one of 10 or so decals specific to the area. Another method I was planning to hide a lot of the geometry lines is to add shrubs in places which are most angular, basically hiding the floor-located deeper lines with a bush. I know that is pretty much cheating, but I don't question it will work 
'>
Now another thing I do right now is separate up-facing faces from not-up-facing faces by a majority direction found in the face normal. If I change from strictly majority, to mostly majority, with a secondary setting added to the function, I may be able to select slightly more or slightly less faces at the position where slope changes faster. This will help to smooth the transition from one texture to the next. Another thing I have thought about doing is using a second part of the original splitting function to backtrack and detect "floating" faces. I think of these polygons in a geology term "xenoliths", which are rocks which separated from one body of rock as molten rock came up from underneath, encapsulating the chunk and placing it in a rock layer it shouldn't have otherwise been in. If I go back and attack these xenoliths, both upward and downward, I can place them more accurately with what type of material surrounds them. This will allow more smoothing to be had at those boundaries.
Back on shadows: while I am interested in finding out the exact details, I am more and more likely to turn them off myself as I work on this. I am much more happy with full quality shadows than I am with the walkmesh quality ones, but I don't know if there is EVER a "perfect" end to that endeavor. I would put money on there NOT being one.
Speaking of the "Causeway", I already downloaded the textures to make that place '> But not right now. I wanted to represent some slowly cooled basalt in the deepest section of my three underdark tilesets which are up next.
For grass on this tileset, I intend to use the same 3/4 human height grass I used in the pictures I put out for the High Plains/Badlands sister tileset. I'm intending to use a high gold - low green grass color, and I suspect the length will cover a great many of the tile seams.
As you have determined, I pretty much don't want the seams to go away on the rocky section, but I don't mind if the other texture is more smooth, I just need to do it in a different way. As you saw yourself when you smoothed the floor mesh, it looks more smooth on that tile, but adjacent to another tile it will look even more sharp-lined at the seam. Not much more because it was pretty bad to begin with on the '_2 tiles, but still more sharp tile to tile. I really need to fine tune those alt tiles more like the way the previous release was. It still had visible seams, but it was much harder to tell which seam went with which tile.
