I get what you mean. For the more tolerant and wasteful-if-at-45º parts like limbs, we export them so it results in the smallest rectangle possible and then just give it the correct orientation at setup.
I suppose I envision it like this:
Some images need to be aligned well relative to another image under it (so it looks like it does when it's painted by the artist). So when you drag an image on top of another image, it helps if it snaps to the same pixel grid. But it seems kind of complicated to have one image have to detect another image under it and handle weird cases like rotations and things, so the one-uniform-grid for non-scaled, non-rotated images made more sense to me. I guess that'd be your call though.
Thinking about it more now, I realize my need actually has less to do with images aligning with the screen pixel grid and more to do with images aligning with each other's pixels.
If it's too troublesome though, I could just continue typing in the values manually. It's not really a deal-breaker. :p
