I think this discrepancy has more to do with how relion_reconstruct generates the half maps, rather than the resolution calculation (I am using the same masks). I recall a popular post/discussion on how cryosparc calculates the fsc with various masks. This may be it, but I can’t recall exactly.
It was pointed out to me that relion_reconstruct does not use the higher order ctf parameters, which could account for the 0.1A discrepancy.