There is something to note, as mike said the telemetry for the battery voltage is only returned every 0.5s and is probably heavily filtered to reduce eronious values, however an intermitent short or voltage drop would not be detected and if you think about the speed of digital electronics, if you drop the voltage below nominal for even 1ms you might start getting odd behavior. It might recover or it might lock up, requiring reboot which unfortunately takes more than 1ms. On top of it, the low voltage error may never be sent if the receiver has already lost power.

I have seen a number of pylon racers install large capacitors on their batteries, to help ensure continuous flow, especially since they try to fit the smallest batteries possible in the planes. Its also common practice in high speed digital electronics to place a capacitor right next to the power leads as close to the IC as possible to make up for the voltage dropof the supply lines during switching/transmitting etc. Although under normal cicumstances you probably shouldnt need to do this, but as chancho said, you have a lot of large servos, and relatively small wires supplying 1 to 3 amps for short bursts.
I'm not suggesting this is a great option, but it should be informative in understanding that what you see on the telemetry or the multimeter is not necessarily what happening, and that battery leads should be short, sized correctly for the current, and connections should be tight.

It does seem odd that you are reading a full 0.5V less out of the battery backer then the battery itself.