Yes, I know several methods how to generate signals or noise at >90 GHz.
The problem is that by using multipliers (US Patent of 1993) or upconverters introduces many uncertainties and the resulting ENR is rarely as good as 13-15 dB used < 90 GHz.
We have used gas-discharge tubes with ENR ~ 15 dB for many years, but they are not too practical now. You need several kV to kick-on the tube, and several Watts to run it.
The people working in THz range use either frequency multipliers or beating two laser signals in a non-linear device. Both have a low efficiency and output power.
---------- Post added at 20:15 ---------- Previous post was at 20:09 ----------
To Lurchman:
I am afraid you have confused two quite different problems: we have discussed noise sources with ENR>13 dB suitable for >90 GHz noise measurement.
The NIST paper of 2002 refers to the correlation measurement of AM and PM noise in oscillators. This is completely different issue.
By the way, I invented the described correlation method in 1966 (M.S.thesis, CVUT, Prage, CZ) two years before Litton patented it in the U.S.