I took a look at this thing from a programming standpoint, and one thing that still puzzles me is why the rotation order is different sometimes. I mean, why isn't it always xyz? So on a copy and paste, sometimes I get xrot, yrot, zrot, or I might get yrot, xrot, zrot, or whatever. I mean there must be a good reason for this (seems stupid to me) but if some lazy coder didn't use the right parsing routine, he might get a zrot where he expected a yrot and just drop the data as a silent error condition. It could be a million other things, but this is the kind of thing that could lose data and not show up in normal testing. I'm not a betting man, but I'll put money on it being a rotation order issue.