For me my G6 always goes up and down much faster than my BG can be in the first few hours (up to 24 maybe). Calibrating in this time period just throws the later days readings off:
Well, that’s consistent 41 high then 43 high. I don’t know how the transmitter handles the calibrations, most likely no one outside Dexcom does (as the ‘uncalibrated’ readings aren’t available.) xDrip+ seemed to use a straight line on the G5 and earlier. A straight line fit to two points with random errors will throw the readings outside the two points way off. Whatever algorithm Dexcom is using internally it can’t fix random errors due to a sensor which is still stabilizing.
I suspect that once calibration has been started it is always better to continue it, so long as our actual blood glucose is stable. Stable means no significant variation over the last 30 minutes (6 readings) because the fingerstick is approaching 5 minutes (1 reading) in advance of the G6.
Ah, well. It’s all spin; the G6 doesn’t say anything for the first two hours, then it spurts random numbers for a few hours, then it runs to 240 hours being fairly accurate. So if you average the errors over time then you get a much lower error than you get over the first day, unless you assume the first two hours errors are infinite
One of the things my G6 does is to suffer drop outs every now and then, maybe once or twice a day. During these dropouts the readings suddenly start descending towards 0, then this stops and they jump back to something like the pre-descent reading. So the transmitter is seeing bad news and, I suspect, guesstimating the right answer. This is somewhat curious.
The obvious answer is to have two G6 sensors running 5 days apart and swapping between them as appropriate. The obvious (no patents please) way to do that is to have the transmitter connect to two sensors either of which can be swapped out as needed.