You may be focusing on wrong parameter. For QPSK, (and higher order), close in phase noise performance is of most import.
For example, TCXO's for GPS receivers (BPSK) have sideband noise specs of:
-45 dBC/Hz at 1 Hz
-75 dBC/Hz at 10 Hz
-110 dBC/Hz at 100 Hz
-135 dbC/Hz at 1000 Hz
-145 dBC/Hz at 10 kHz
-150 dBC/Hz at 100 kHz
This is for their fundamental mode TCXO at about 14 to 20 MHz, used for synthesizer reference. To be fair, GPS needs exceptionally good close in noise performance to maintain correlation lock on a very weak CDMA signal. (-145 dbm to -155 dbm at receiver input). The GPS TCXO also has 0.5 ppm frequency accuracy over operational temperature range which is necessary to speed initial time to first fix performance.
High phase noise will result in the phase modulation constellation to be 'fuzzed up' with jitter which will degrade low signal level recovery performance, whether the jitter is created in the transmitter or receiver side.
The parameter of 'fuzz balls' in a nPSK constellation signal is called Error Vector Magnitude (EVM) and is specified as % peak and r.m.s. error over some period of time.
Agilent and Rhode and Schwartz test equipment supplier's web sites have some good application notes on this topic.