Is linearity a problem in these sorts of devices? I'm not sure where it would come into play. I'm assuming this does IF processing in software and it doesn't have a classical detector where linearity would be a problem.
Yes, return loss and directivity of the bridges/directional couplers in there is a big factor. It directly corresponds to usable dynamic range after calibration. If your directivity is only 20 dB at 6 GHz, 110 dB of dynamic range at the detector won't do anything for the quality of the final measurement. So an honest way of specifying dynamic range would be for a fully calibrated device. Giving that they say it has 110 dB, I don't think that's the case here.