I was not involved with the calibration or testing of the mobile W-band radar since it was built at the U. of Massachusetts. Bluestein and Pazmany published a paper on the engineering of the radar itself in BAMS in (I think) 1999. The AMS Journals website is down right now or I could get you an exact reference.
I operate on the assumption that the data I work with is true, or at least true enough to be useful (which is what really matters). Obviously, you always have to take the velocity measurements with a grain of salt. As with any Doppler velocity data, it could be contaminated by a lack of scatterers (i.e. in the torando's eye) or a really big scatterer (like a bird) that biases the power spectra in each pixel. I tried to account for this by threshholding the velocity data on low reflectivity (-18 dBZ or less) and manually editing out a few pixels of obviously bad data. After those filters had been applied, there were still enough points with a velocity of 50 m/s or greater left in the tornado that I believed those velocities were real. (Tornado threshold velocity is 35 m/s, and we know for a fact that there was a tornado in the sector.) GBVTD analysis showed that those velocities existed all the way around the tornado funnel during its mature stage.
I hope that answers your question. Radar meteorology is a much more inexact science than we'd like to admit, but it's still immensely useful!