Some suggestions as to how this could be solved (which I can't check now since I'm at uni)
- Where does this matrix get read in in the source code?
- What variables are linked to this data?
I'm pretty sure that if we tracked the use of the matrix through model loading (and if entirely necessary, through physics) we'd either figure out how it does what it does or at least be able to blindly reverse engineer the solution.
I'm willing to do this, but I need some more information on this since I've been out of the loop for a while (and specifically, didn't actually know this was a problem).