Here’s a thought - If PG&E is like some other utilities, it will attempt to fill missing smart meter readings with so-called “calculated reads”, where PG&E looks at the energy gap between the meter reading before the smart meter stopped transmitting, and when the meter started re-transmitting, then tries to fill the gap with “calculated reads” data that sums to the difference during the gap. In your case, that might mean that the total usage during the last incomplete 15min period before they took out the power and the first incomplete 15min after the power went back on, got spread over the 6 or so hours that the lines were disconnected.
I’m betting on the “calculated read” theory. But unlike NV Energy, PG&E doesn’t alert us to calculated vs. actual reads. At the same time, based on my Sense vs. PG&E data, I’m pretty sure that I’m seeing 99.9% actual reads fro my PG&E account vs. lower numbers for NV Energy (sample size 2 - I have seen a low of 15% and a high of 98%)