To minimize the effects of interpolation, it would be better to concatenate transformations including the motion correction transformations. You can align the EPI to the anat of the same session and then align that anatomical dataset to the first session's anatomical dataset. Alternatively you can align the EPI dataset to the the anat of a particular session or to an in-between anatomical dataset. align_epi_anat.py includes the second option, and the other methods mostly make use of align_epi_anat.py which then calls 3dAllineate. You can see some examples of aligning across sessions in the class material (slides 25,26) and in the link below.

https://afni.nimh.nih.gov/pub/dist/edu/latest/afni_handouts/afni10_volreg_talairach.pdf

https://afni.nimh.nih.gov/sscc/dglen/alignmentacross2sessions