Just to add some confusion and something else to think about.... when doing static testing when I built my power buss that I use, I found just the opposite to be true, running the system at 4.8V it consumed greater amperage than running the same tests at 6V, therefore, equal capacity packs of 4 and 5 cells under static conditions the 5 cell should take longer to draw down to "cutoff" voltage.....I did static testing by using servos loaded to half rated torque. but this was using my buss and deans connectors replacing the "stock" servo connectors,and dual 10amp capacity regs setting the voltages while using a 6 cell 3000mah pack as the supply.

Since I haven't run the system in the air with different voltages I can't say if it still holds true under actual flight conditions.

However I do agree that with the typical "stock" setup your observation is accurate.....kinda interesting I think....
As Dick pointed out, a servos speed under no load is not = to speed under load, and servos will try to run as fast as they can, maybe the servos are sending you a message.....give me the electromotive force I need under load and I will reward you with closer to spec performance. in other words, considering the large voltage losses I have seen in testing across "stock" connectors, they are acting as restrictors, with the higher pack voltage the losses are slightly less so the servo now can operate with a voltage closer to what it really needs. but at the expense of shorter pack life.