I also see the mode change issue with the virtual sensors - what I think happens is that if you've set a mode (eg. average) on one sensor, then click to edit another different virtual sensor (eg one that should be "highest temperature"), the first mode set (average in our example) carries over to the newly selected sensor, but the fact that it has done so doesn't actually show up until you've left the page and come back later. I always sanity check the results if I've changed anything because of this issue.

Agree that your calibration method is as accurate as you can get - for all practical purposes it should be basically perfect. One possible (small) caveat is that last time I recalibrated, I noticed that my water temperature sensor readings changed a bit (maybe a degree) immediately I started my pumps, the water flowing over them presumably helped to stabilise the temperature in the sensors, even though everything had been left to reach equilibrium for a long time before starting the process. It happened so fast it couldn't have been heat dump from the pump (actually, I can't even recall if it was an increase rather than a decrease anyway), so I went with the calibration value with the pumps on rather than off. I guess it could be friction from the water passing through the tubes and sensors (if it was an increase).

Of course, if you run the pumps for any appreciable time before calibrating, you'll have a bit of heat dump from them so the water should read a bit higher than the air at that point, so for best accuracy I'd calibrate first with the pumps off (so everything is really close) then turn them on and quickly alter the water sensor calibrations to get them really nailed.