Dear Doug et al.,
a while back there was a post on peculiar registration problems using reg-feat2anat, e.g., the registration using reg-feat2anat was quite off (seriously rotated or similar), whereas barebones flirt on the same volumes did better. I didn't spot a resolution on the message boards, but I have found that the EPIs end up being rotated incorrectly (around A-P axis) when I run flirt with the initial registration (from tkregister2-cmdl), but not with the same command minus -init. I didn't see any hidden option flags in reg-feat2anat for blowing off -init (although did spot one in fslregister), so am guessing that other people haven't found this to be a problem.
I should mention that I'm using reg-feat2anat on a partial volume scan - there may not be the same issue with whole-head EPIs...
Apologies if I missed a previously posted solution,
best, Fred