OK, I hadn’t come across these before. Their b=0 is clearly quite a bit away from zero, even though they state that they acquired b=0 images in the release notes. I’m not sure what would cause such large values here, the 3T HCP data has b=5 for their ‘b=0’ volumes. Can you double-check that the original bvals file lists the same value for the b=0 volumes…?
In any case, the simplest fix here is probably what I’d suggested in this post.