Hrm, that is unexpected.
Well the problem is, that the normals are just fine. The preview of the original and exported GR2 is a nearly 1:1 match:
https://dl.dropboxusercontent.com/u/32263228/dos/gr2/orig_export_comp.png
https://dl.dropboxusercontent.com/u/32263228/dos/gr2/orig_export_tangents.png

I suspect the screwups are because tangents are calculated differently (the only difference between the new version I sent you and the original is that I changed the direction of the binormal). These are not exported however, so they must be recalculated when converting to GR2.

Which model are you trying to modify btw?

Last edited by Norbyte; 21/01/16 09:42 PM.