It is odd for DELTA to be bad when 3dinfo correctly shows
2mm voxels. Would you be willing to upload those datasets?

Regarding the orientation, you might want to run 3drefit
-orient RPI on them, so AFNI programs know which orientation
they are supposed to have.

It is disturbing that FSL is still creating datasets with
sform_code = qform_code = 0 in them. That defeats the entire
purpose of using NIfTI (to define the data location in space
between packages). You might even consider asking them about
that, but it is of course up to you.

Anyway, if you would be willing to upload the data, I'll try
to see what is going on with the DELTA values.

Thanks,

- rick