The duporigin option only resets the header to make the origins match but does no alignment. It is useful for those cases where one already knows the data are aligned and needs to force the dataset headers to match. If data are acquired in different sessions or at different obliquities, you will need to align your data together. Take a look at this page I just wrote up:

[afni.nimh.nih.gov]
[afni.nimh.nih.gov]

It may be difficult to see differences in alignment without seeing some examples first. For some more info on visualization, see this link:

[afni.nimh.nih.gov]
[afni.nimh.nih.gov]

Yes, all data can be oblique, and there is no real limit of obliquity. It seems at least one manufacturer has made it the default to be slightly oblique. More than 45 degrees will make it hard to tell the rough orientation though. Usually one wants to align data by either matching the obliquity (all sharing the same oblique angle or all cardinal) or aligning the data (align_epi_anat.py includes obliquity together with other affine transformations). Once data is aligned however, the reported degrees from plumb are invalid and should not be used and can be removed with 3drefit -deoblique. For other information on obliquity:

[afni.nimh.nih.gov]
[afni.nimh.nih.gov]