@Rick: I've tried with a zpad of 500, the largest my machine would allow, but I still have the same issue, in this case that the superior part of an epi that was originally centered "below" the structural dataset, is cut off in the output. Perhaps I should mention that in this case, and in fact in most cases, rotation is actually not necessary for alignment, so I only do translation, like so:

3drotate -linear -noclip -zpad 500 -rotate 0I 0R 0A -ashift 20S -5L 20P -prefix output+orig input+orig

could that be why the zpad option is not helping?

@Daniel: Yeah, that sounds good. How do I upload the data? Even if you can help me fix the alignment, however, I still think the underlying issue would be worth addressing.

@Isaac: It seems like what you are suggesting could potentially resolve my problem. I usually refrain from messing with headers manually.