I have run a bunch of cases through bbregister and overall it worked great! 315 out of about 325 cases
the alignments were very good, but on 10 subjects it did very poorly. I can not so far find anything unusual
about this data. Per earlier posts, I have tried the different initilization options with no luck. I also tried an
fsl_rigid_register 6 DOF to create a starting .dat ,but that was way off too, maybe because it
uses similar initilization information? I also saw the post to use --init-reg-out to save the init transform
but I am not sure what to make of the output.

This is aligning data that has been through the 1000FC pre-processing. This converts our RAS
scans to RPI, and also de-obliques using AFNI. Would you suspect either of these steps as
confusing bbregister? We are using dcm2nii (mricron) for dicom to nifti conversion.