According to Jared's video those capacitors charge up to about 0.3V and are compared to the divided battery voltage. Of course there is also the 1uF caps that would then also charge to 0.3V, since they are all in parallel.
I also note that those caps are much higher in value than the 2pF and even 2nF caps that you are thinking of using.
My other thought is why not use a much simpler circuit without the batteries? Since you are just doing calibration, why not just put a DMM on the actual peak detect voltage and skip canceling out the 1.2V? Then just calibrate to keep this voltage constant, no matter what it happens to be. This is way more general purpose. This would eliminate the batteries and the circuits required to cancel out the 0.3V levels, as well as the calibration becoming invalid as the batteries wear down. And it would eliminate fully 1/2 of the device, or would you still need to rectify both polarities of the input? Or what would you get if you left both polarities in and put the meter between the + and - outputs? Shouldn't they just cancel out anyway unless the positive and negative halves of the input aren't symmetrical?
