I have a new theory - perhaps those are more granular instantaneous consumption rates. If you look closely, all of the blips drop below the main line. Perhaps those are times you were accelerating from a stop or at least not running steady state. The main bar chart is obviously a trending average. For example, right around the 1 hour mark there's a blip on the lower line and the big bar is relatively short. There isn't another blip until about 3/4 of the way to the 2 hour mark. All of the big bars are trending upward during this time. Looking at the blips and bars during the first hour also seem to have a similar correlation.
It should be a pretty simple matter to test this theory. Especially if you shorten the timebase to something way less than 2 hours.