Calibration lines that are continuously injected in the interferometer are used to estimate time varying factors of calibration model which are in turn are used to correct h(t) in GDS pipeline. Since the calibration lines are of finite height, whenever there are strong glitches in the interferometer data the estimated time varying factors tend to be corrupted. To avoid these glitchy times, a threshold is used on the coherence based uncertainties at calibration line frequencies (using coherence between injected signal and DARM error).  For C00 and C01, at LHO we used a maximum threshold of 0.4 %. The attached plot show the distribution of estimated uncertainties at various calibration lines during all of O2 (only observation intent times were used). We see that 0.4% is a little bit low, especially for PCal line 2 (331.9 Hz). From the plot, it seems any value above ~0.6%, outside the peaks of the distributions, will work fine. But we also don't want to go too large otherwise it would defeat the purpose avoiding glitchy times. At LLO we use a threshold of 2%. Since 2% satisfy our requirement of > 0.6% and also makes the thresholds between the two sites consistent, we could use 2% threshold for LHO also (for C02).