The above mentioned methods are incomplet. Two terms of CMRR exist: systematic and random. To simulate random CMRR you must perform Monte Carlo analysis as random CMRR arises from mismatches between signal paths in your amplifier.
This is a very common "newby error". When you simulate the CMRR of an amplifier without considering the mismatch effect, it is common to obtain CMRRs of more than 90dB. Now, when you add mismatch, the CMRR substantially decreases.