+1 what armians said. Also, don't subdivide your mesh all the way up before projecting for the first time. Project for each subdivision successively starting from the lowest one. That way the shape change isn't as drastic and, hence, the algorithm is more likely to perform as expected.
Hm you're right, it seems to have problems with single faces here too. My first test was projecting on 3. Two are fine as well, maybe the curve split algorithm needs to touch at least 2 faces to work properly for some reason. Wonder if they fixed this problem in 2013.
If you can't do any of warby's suggestions, another alternative is to change your viewport rendering to Viewport 2 and change the Transparency Algorithm to Depth Peeling, though this is strictly for Maya and won't fix the extra in-game draw calls involved with having a transparency shader on a model that has no…
I didn't see it mentioned as I read some of the comments but is there the possibility of variants? Example: Instead of having - Orange, Orange Orange Can you make - Orange1, Orange2, Orange1 To break up the potential repeats in long straight areas? Or does the algorithm try to avoid these types of situations.
MoP: It's the depth of the vector, not the surface. The vector always has length 1 since the lighting algorithm expects a normalized vector. It gets darker the more the bumpmap normal deviates from the surface normal. Two parallel surfaces will have the same blue value no matter what their difference in depth is.
@Japhir - magic of google's algorithm at work. It's more relevant than the Public Service ads because it is actually targeting something relevant to your location. They only have so many game related ad providers so at some point they go for the easy local sell.
A Master substance thread is linked below this has a lot of great examples or links for substances. It keeps on growing and a group of very knowledgeable people stick around to answer question including people from algorithmic Substance Designer. http://www.polycount.com/forum/showthread.php?t=129560
The humus article talks about tessellation. And there are several tessellation algorithms. If the model is not going to be tessellated, any method is valid. I use to use B in certain models (it's a way to have quads in a cylinder of parity 2), but for normal mapped models A gives me better results.
I'd concentrate on the sub-surf first and then do another detailing pass. I use Blender and the raw base vertex shading algorithm is hideous. Adding a sub-surf with some edge-defining loops will make your bolter look 100% better straight away.
If the base workflow is developed its getting harder and harder to refine. I agree that better uv algorithms and retopo tools will help... And im shure a lot of clever people try to develop such tools... Its like the wheel... Try to find something better to move a car...