These cheaper SA's etc ike the DSA 815 from Rigol appear to have frequency varying impedance at the input. Thus unless you're using the calibrated tracking gen as the input to your DUT you're going to get some dB variance from your true signal across the range. Thus if you have a signal generating device, VCO or whatever that you're characterizing, there's going to be an instrument accuracy issue at some frequencies.
Which leads me to the concept of a device which will provide decent calibration for DSA's which implement correction tables.
A device/VCO/sweep CW which connects directly (no coax) to the DSA and outputs a flat amplitude (+/- 0.1dB perhaps) across the required Freq span as driven by a USB control signal from a PC/laptop.
this allows the user to implement & save the DSA correction table to get the flatness optimizations.
Such a device can also drive a DIY Lo z probe and also allow for another correction table to compensate for BOTH the probes and the DSA VSWR issues.
Interesting?