In trying to post-project fill-in data when the museum barometer died on 20 Aug, I see also odd spikes/level shifts on 19 and 22 Aug. On 20 Aug, Pcell seems to be pretty good, though high by about 0.2mb. However, I can't justify this 0.2mb shift looking at the entire 6-day period. Thus, I've decided simply to substitute Pcell as measured for all of the museum P values.