I did have the battery check happening just before going into sleep mode, so I guess it makes sense that in some cases the battery would be low and trying to recover. I'm assuming the peaks I was seeing could have resulted from when sensors were checked and there was no change in value and so no transmit which would have left the battery with more juice.
I've moved my battery check routine to be the first thing in the loop and I'll see how it goes.
@Anticimex, your point about tuning the algorithm is a good one because I've been using NiMh batteries to power the system which are topping out at about 2.8V when fully charged, and the booster is supposed to work down to 0.8V, so I'd like to realign the range for the reporting to fit this. Any suggestions on the best way to do this?