Issue is that HP8568B is not enough good for craracterize this phasenoise.
Test setup:
SDG1010 (1Q13) With HW:02.... and FW....27
Sine, CH1, 50ohm, level 0dBm
Internal frequency reference in use.
Small (under 12ppm) frequency offset corrected in phasenoise software what use HP8568B via GPIB. Software is KE5FX
Frequencies have checked using 53131A and GPS OCXO
HP spectrum run (of course) using its own very good OCXO)
Result was so good for this kind of function generator that first I think there is something really badly wrong.. but after many many checkings I keep this result reasonable valid. More than 1kHz offset I believe that result is better but here really specrum limits. So there is "phasenoise issue" with this test method with these equipments.