About the Agilent VNA Uncertainty spreadsheet, a couple of questions:

The sheet appears to yield higher-than-expected uncertainty magnitudes. For example, uncertainty of S21 that is calculated with the sheet (for measurements of normal, not pathological, devices) appears to depend little on the VNA model,the cal kit selection, averaging or video bandwidth, or initial estimates for S21 and S11. There is a dependence on frequency band, of course, so let's say we are measuring at 2.4GHz, for example. The lowest uncertainty magnitudes in dB fall between roughly 0.10dB and 0.15dB, which seems larger and it seems more irreducible than I expected. Perhaps I missed it but are these 2-sigma numbers? For no particular reason, I would expected less than 0.05dB under optimum conditions with the best VNAs. Am I operating the spreadsheet incorrectly? Is there detailed background available for how this sheet computes its results and some guidance on selection of parameters?

The sheet appears to yield higher-than-expected uncertainty magnitudes. For example, uncertainty of S21 that is calculated with the sheet (for measurements of normal, not pathological, devices) appears to depend little on the VNA model,the cal kit selection, averaging or video bandwidth, or initial estimates for S21 and S11. There is a dependence on frequency band, of course, so let's say we are measuring at 2.4GHz, for example. The lowest uncertainty magnitudes in dB fall between roughly 0.10dB and 0.15dB, which seems larger and it seems more irreducible than I expected. Perhaps I missed it but are these 2-sigma numbers? For no particular reason, I would expected less than 0.05dB under optimum conditions with the best VNAs. Am I operating the spreadsheet incorrectly? Is there detailed background available for how this sheet computes its results and some guidance on selection of parameters?

The only 2 sigma numbers are the residual systematic errors (directivity, reflection tracking, source match, load match, transmission tracking). The rest are all worst case numbers. Additionally the numbers are not RSS'd but added worst case. It is hard to comment on your results without knowing the specifics. I just looked at one of the plots for the N5242A using an ECAL. The transmission error was between .05 and .06 dB for low losses.