I tend to agree that it is not perfect and quite limiting taking into account those two features. Question is, is it possible that with the data at hand, you could still produce something decent. Maybe some kind of random or calculated variable intensity ? The blending could maybe be fixed by assigning two textures to the same landtype and then draw them both at the same location. I'm just throwing some stuff out here to see if it would be manageable.

I created a new binary with the l3dt code in it as well now.

http://users.telenet.be/sunspot/dos/DosTerrainImporter.7z

Hope you like it.