Ok, one last post here - I've got it working very nicely now. Here's what I did, hopefully this can help you Michelle, at least for future subjects. I didn't need to change my acquisition, it works ok with what we're already doing.

Scans:
1) high-res MPRAGE T1 anatomical, 1mm isotropic, sagital acquisition. This scan is called anat+orig. Acq time 6min with 8-chan coil, ASSET parallel imaging.

2) FSE T2 - 38 slices, 3.8mm thick, .438mmX.438mm inplane res, axial oblique slices, aligned to AC-PC line (as seen on the previous scan - that's why I acquire it sagital). Acq time: about 1min30sec. I do this scan right after the previous anatomical. So there is very little movement between them, and coregistration is not necessary.

Then, after importing the scans into afni:

3dSkullStrip -prefix strip_fset2 -input fset2+orig
(strips the fset2 image, which usually works well even in stroke brains).

3dWarp -oblique2card -gridset anat+orig -prefix cardstrip_fset2 strip_fset2+orig
(regrids the oblique fset2 image to the spacing of the T1)

3dcalc - a cardstrip_fset2+orig -b anat+orig -prefix stripanat -expr 'step(a)*b'

The result is a T1 anatomical pretty nicely stripped. There tends to be some eyeball left, and a wee bit of dura, but it's much much better than trying to strip the T1 directly.

-Jed