Hi Robin,
Okay, I will test with the NL_warp option.
Regarding the anat resolution, it goes down to 2 mm^3 because you are using a 2 mm template, MNI_avg152T1+tlrc.
Importantly, that is a low-res blurry template, and as such, we would never recommend it for non-linear registration (or even affine, for that matter). NL-reg should use a fairly high-quality template. The current practice is to actually run @SSwarper (to align to the MNI152_2009_template_SSW.nii.gz template) for skull-stripping and standard space warping *before* running afni_proc.py. Then the output of that can be give to afni_proc.py via -tlrc_NL_warped_dsets. See example 6b:
afni_proc.py -show_example 'example 6b'
But anyway, I will run this with -tlrc_NL_warp and let you know how it goes.
- rick