nicreap-

afni_proc.py (command line) and uber_subject.py (graphical user interface; GUI) are the "newer" superscripts in AFNI that will handle all of the data preprocessing and single-subject analysis, including (but not limited to) slice timing correction, motion correction, coregistration, normalization to a template, smoothing, scaling to percent signal change, and regression. As part of afni_proc.py, align_epi_anat.py is called with recommended options and I suggest everyone give it a try.

My experience with alignment is that you're going to get EPI interpolation one way or another (happens in slice time correction, motion correction, coregistration, and normalization among others). The challenge is minimizing that interpolation, which afni_proc.py/uber_subject.py attempt to do by combining several of the transforms. You can try to align the anat to the EPI using align_epi_anat.py (-anat2epi, also the default), and let us know if that works better.

Are you using any type of censoring in your 3dDeconvolve step to censor TRs containing high amounts of motion?