I reported in a comment to alog 85961 that I pushed a new calibration today based on report 20250719T225835Z. However, I made a mistake: I did not load the CAL CS model after this push. This load includes changes to gains and filters in the CAL CS model that are important for the calibration. We went back into observing because an unloaded model does not prevent us from observing, and I was not aware that I needed to do this! I found out about this mistake when Dave brought the unloaded filter changes to my attention.

Overall, the differences were small, but they were clearly enough to make a difference. With permission, we left observing, loaded the model, and TJ ran a broadband PCAL measurement. I have attached a comparison of the three broadbands from today. Green is the failed push of the 6/28 report, blue was what I thought was a successful push of the 7/19 report, and then red shows the result after we loaded the model.

Therefore, the 2% uncertainty at 20 Hz that I noted in my previous alog was due to having the wrong filters in CAL CS, and we don't have any problems with the L2/L3 crossover.

So, as a note to myself for the future, make sure to load the CAL CS model before observing.