I am guessing here, but I would imagine that including -anat_has_skull no prevents skull stripping from being performed. You are then performing non-linear normalization to the 2009 MNI template, with no skull. That is definitely going to cause problems for the algorithm trying to match the images, and that is my guess for what went wrong.
I've been very fond for using @SSwarper, the AFNI program that performs skull stripping and normalization all in one go, with some tweaks and optimizations. You can use that on your SurfVol in the SUMA folder, and then import the needed files in AFNI proc. This is very useful as you only have to run non-linear normalization, which is slow, just once - so if you decide to run afni_proc again, and again as you tweak settings your life will be much more pleasant.
I've found this recent AFNI publication extraordinarily useful:
link to post.