So far, all of the observations for the H2CO densitometry project have been performed at the Green Bank Telescope. During a 10-day long observing trip here, I've learned a lot about diagnosing bad data.
This first image shows TSYS vs Airmass for good data. The high outliers are just sources with continuum in them - the continuum is the source of the extra signal, not atmosphere. The receiver temperature is a nice 20.6 K, and you get about 5 K extra per airmass, suggesting a zenith optical depth of 0.018 assuming a round atmospheric temperature of 300K.
In the same style plot, there is a set of observations with low system temperatures: that stuff is good. There is also a set with clearly rising system temperatures, even at constant elevation. These data are bad. During this observation, the "blowers" that are meant to keep dew off of the receivers failed. Dew buildup on the receiver covers lead to higher optical depths and therefore system temperatures.
Finally, this data set was totally useless. Ku-band is not particularly sensitive to water in the atmosphere... but it's still not a good idea to observe during a rain storm. Note that the fitted receiver temperature TREC is nonsensical.