Hi, Korann-

A few things there.

What is your AFNI version ("afni -ver")? If it is quite recent, then you should get a directory called QC_[something] (in your case, probably QC_INSP004) that would have a full report of quality control (QC) blocks from your afni_proc.py processing. This is described in detail here:
[afni.nimh.nih.gov]
It will show you individual alignment step results (EPI to anat, and anat to template), and you can get a sense of what piece might be running amok.

For aligning to MNI 152, we have been recommending that people use @SSwarper as a pre-afni_proc.py step generally. Basically, you provide your raw anatomical volume, and this program will do bth the skullstripping *and* estimation of a nonlinear warp to standard space. You then provide the results of its hard work to afni_proc.py. If you read its help file, then you will see how to do both steps:
[afni.nimh.nih.gov]

@SSwarper          \
        -input    ANAT_VOL_WITH_SKULL      \
        -base   MNI152_2009_template_SSW.nii.gz        \
        -subid  INSP004        \
       -odir   SOMEWHERE       \

There will be some snapshot/JPGs created automatically by that program to show you how the alignment (and skullstripping) both look.

... and then your afni_proc.py command would contain some lines like the following (just grabbed this from @SSwarper's help file, where ${subj} is the subj ID and ${tpath} is the path to the reference MNI vol)

    -volreg_tlrc_warp -tlrc_base $tpath/MNI152_2009_template_SSW.nii.gz              \
     -tlrc_NL_warp                                               \
      -tlrc_NL_warped_dsets                                       \
         anatQQ.${subj}.nii                                       \
         anatQQ.${subj}.aff12.1D                                  \
         anatQQ.${subj}_WARP.nii

That is how afni_proc.py gets the results of @SSwarper.

While I see that you are not censoring data, I will note for future reference that, f you do start censoring your data, you should read this thread about how best to calculate RSFC params:
[afni.nimh.nih.gov]
Basically, censoring complicates things- as does bandpassing, to some degree (and to some degree of freedom, but that's a separate note).

--pt