Hi Rick,
the problem is that they did not supply any head motion estimations for the 7T data at all.
However, after checking the data more, it looks as if their so-called minimally preprocessed data already included a perfect motion correction.
When I run through the single volumes, the brain looks almost perfectly static, i.e., there is almost zero motion.
This is what they state in their manual:
Quote
FLIRT-based motion correction using the “SBRef” volume as the target (Smith SM et al., submitted).
I just didn't expect that their motion correction would be so good. Maybe after all motion correction or even censoring in AFNI is not required.