I suspect the differences would be quite small. Typically EPI images are collected in the axial plane and anatomical (T1) are collected in the sagittal plane. My experience is that in general one does not get noticeable improvement by doing alignment to the in-session anatomical vs. another anatomical of the same subject. There are exceptions to this, but using an average anatomical or template anatomical should work just as well if not better for your purposes.

You can certainly use AFNI to align two anatomicals and then average them together. You could use align_epi_anat.py with the -dset flags.

The process that I take is to use Freesurfer, which allows you to either average two T1s together (recon-all with two input flags) or to do a longitudinal model that will make a template from the two scans.