The rotation info file used with the deformation data contains the axis as defined by two points. In my version of maketarget that included support for rotation and multi targets (makepose?), I found it limiting to define the axis by only two points. What I came up with was being able to define the axis with two sets of points. Each set of points can contain one to any number of verts. Then finding the centers of each of the two groups of verts defines the axis. This was more flexible in placing the axis exactly where it was needed. This was my approach when the base mesh had an actual skeletal system. The placment of the polyhedrons within the new mesh suggest that you are using those to define the joint axis and rotations. Can these structures themselves be deformed to define non-perpendicular axis?
Secondly I added a third file to the rotation definition. This file was an extra set of deformation data to be applied during the rotation. It allows for example, the bunching or stretching of skin around the joint when the joint approaches the maximum abductions and adductions.
Another inovation I was looking at was being able to allow deformations to be applied to the posed object. This doubles the calculations and therefore the calculation time, but I think would be a very usefull feature.
I'm looking forward to the next release and hope I can contribute in some way.
Tim