[Freesurfer] skull strip problem

Hello, We have a subject processed in V4 for which T1.mgz looks fine (white matter ~ 110; no obvious artifacts), but the brainmask.mgz is basically all zeros (except for a section of the background in slices 200-255 of the coronal view).

The recon-all.log file is attached.

I don't know if the problem is with the derivation of talairach_with_skull.lta that is used in 'mri_watershed' in V4, or if the issue occurs in 'mri_watershed' itself. (NB: There are a lot of "ambiguous basin" messages as part of the 'mri_watershed' output).

How do you check the accuracy of talairach_with_skull.lta? Any other suggestions (particular options to change for the watershed...)?

thanks, Mike H.