[Freesurfer] cortical mask for group analysis

I would like to make a mask for group analysis that excludes the unknown parcellation area. Since I would like to apply the same mask to all subjects (I have already brought them into alignment with mri_preproc), I was wondering how I can convert the average template label/?h.aparc.annot to a .mgz files that I could then convert into a mask. Is there an alternative way to produce such a mask?

As a follow-up question, I was wondering if parcellation areas such as the corpus callosum should also be excluded from analysis? Are the thickness estimates made in the corpus callosum parcellation area interpretable?

Thanks in advance for your time.

Cheers,

-Morgan