For the sake of it being Friday afternoon, it being easier, and also the method that I use in my lab, I would recommend option 1.

The output of the mris_convert command is transformed to the fsaverage space. So there is no need to create standard meshes. Standard meshes created using MapIcosahedron are another way to get all of the data ready for group analysis by getting vertex correspondence. Since your data is already resampled to fsaverage thanks to -qcache and mris_convert, there's no need to do anything else. There's some bits of this in the SUMA.pdf documentation, and possibly in the new documentation at [afni.nimh.nih.gov]

If you use option 1, which I would recommend, you do have to perform a t-test on each hemisphere separately. This is standard practice in Freesurfer analyses (LH, RH, and subcortical being separate analyses).