afni.nimh.nih.gov

The default cost function, lpc, and several others seem to work well to align epi and anat data, so I suppose you're not aligning the same two things. I used the first sub-brick of the EPI dataset (sub-brick 0). That sub-brick is pre-steady state and has the best contrast. Skullstripping did cut a little off the top, but it did not affect alignment at all. To use the pre-steady state sub-brick as the alignment base in afni_proc.py, consider using the -align_epi_ext_dset option.