That's right, Rick. It was all processed within AFNI, using the programs you listed. I did notice what I thought was a mis-specified time field, but I wasn't sure if that was what was causing this issue with converting to NIFTI, or how to fix it. My intuition is that the 3dbucket step when I re-order the sub-briks across subjects to all be in the same order is what created the problem.
In case it's not clear from the history in the header, here's the command I used for that:
3dbucket -fbuc -prefix ${subj}.REML_IMs_reord ${subj}.REML_IMs+orig'[1dcat reorder.1D]'