I am having same problem with results difference when comparing genesys momentumGX results with CST MWS results.
I suppose this is because of meshing types used in momentum and CST (hexahedral mesh type) are different. In my case, I simulate coupled microstrip filter in a hairpin topology, with 75 degrees angles, so edges of lines are not parallel with CST generated mesh, this should give accuracy loss in CST.
However, I was attempting to make finer mesh into CST, and still have about 4dB S[1,1], S[2,2] difference between CST and genesys...
Would appritiate any help about why does this difference appears.