Precise Mass Summary Output
One of the many checks one can do in FEA is to compute the mass, inertia and center of gravity. Besides doing this for the normal models, it can quickly extract out the mass and inertia of the superelement (SE) to verify the MATRIX50 that you're bringing in has the same mass and inertia as the genuine article.
If the mass comes in too low, try specifying more modes in CMSOPT to see if it fixes it. My previous co-worker also came across intermittent issues with using constraints equation (CE) inside a superelement. He replaced it with really stiff beams instead before generating the superelement.
Personally, I would take a bit of time to select master nodes (M) not just at the connections but at various points on the model to properly capture the modal mass. This would be similar to identifying accelerometer locations on your part to properly identify modes of interest for a modal test.
Here's the command to calculate it all :
ALLS $ /SOLU $ IRLF,-1 $ PSOLVE,ELFORM $ PSOLVE,ELPREP $ IRLIST