[thegreatduck]

So I was able to find a model with the bones and compare it to the model on modeler's resource. The problem is this. Mario's points in that model are on an integer grid with an interval of 0.047 or something like that. The rotation points are not. Computing the transformation from one submodel to another is messy at best and wouldn't give me whole number matrices that wouldn't introduce error. So one solution would be to find the magic linear combination of points in each submodel that exactly add up to the rotation point and then do the same for the MR model. Unfortunately this is just too difficult. So the obvious solution is to just use the other model I found which had bones in it. No idea on Bowser though. I'll keep digging.