that's what I would guess, but why not smooth out the data then with the average of the two days on either side--or at least put an errata note that it's not accurate... makes it look like the network collapsed for a day.