I will add that align_epi_anat.py does not reorient from RAI to RPI by default and preserves the original orientation because there is really no need to. There are various "-master_" options that can be used to override the default behavior if you want something different. The AFNI viewer shows the datasets in underlay or overlay as they are stored accommodating for the origin, orientation and voxel sizes. So if an EPI dataset is placed over an anatomical dataset, it's not required that both have the same orientation.

Obliquity is another complication we're considering (oblique/deoblique overlay to match underlay) and might possibly be included in some future version. For now, the standard simple orientation (RAI,LPI,...) is stored separately from the matrix that defines an oblique dataset. The viewer treats the datasets mostly as rectangular datasets for the "orig" view without tilting the dataset one way or the other. With 3dWarp, you can "cardinalize" an oblique dataset or "obliquify" another dataset to match. See this page for more info:

[afni.nimh.nih.gov]