I'll agree with Peter that your problems are likely caused by the fourier interpolation that often gives ringing artifacts. Switch to -cubic instead and see if that works. There's a fair amount of ghost artifacts in your data that we don't typically see. Still, I don't think a rigid body correction scheme will work as well as an affine one. See this previous posting on using align_epi_anat.py to do this with 3dAllineate and affine transformations.
http://afni.nimh.nih.gov/afni/community/board/read.php?1,68951,82860#msg-82860
Even better, try the Tortoise package to handle all this for you.
