About drivetrain loss, here is some good info I have:

"There are two main causes of power loss in transmissions.

1: The nice simple textbook loss due to sliding contact of the gears. Taken to be constant with speed, typically about 2% of power for spur gears and 3% for hypoid bevel.

2: The complex loss (never found in textbooks) due to oil fling and windage. As the gear starts to rotate it picks up oil, has a large wetted area and the loss follows a normal V^3 drag power law. As it picks up speed it tends to fling the oil and carve a groove in the oil bath reducing drag by entraining air. As it flings oil the oil depth reduces, again reducing drag. It moves to a loss approx proportional to speed regime. I don't know what happens at very high speed when the oil level has been reduced as low as it can go or a larger gear on the same shaft is still flinging oil and a smaller gear runs clear of the oil bath. Auto boxes main loss is due to pumping loss in the oil pump and hydraulic system as the gears are not dipped in oil.

Increased temperature will reduce viscosity and reduce drag.

My LSD diff has an oil cooler, not for the benefit of the gears but to maintain the oil at the working temperature of the Viscous LS unit. It has a warning light to tell me when it is possibly an open diff.

Transmission loss is a black art known only in the makers dyno house and consultants like Ricardo or SwRI, there is very little published information. Lots of people quote some % figure for loss but don't say what the power or speed was so the figure quoted is useless. Even assuming it's for peak power and speed doesn't tell you what it will be at normal cruise. In lots of engineering texts I have only ever found one graph of transmission loss against speed, it was in a book on dynometer testing of engines, 1936 reprinted in 1969!"

and

"Losses in hookes joints and cv joints tend to be small, and pretty much proportional to power.

Losses in gearboxes tend to be rather larger, and tend to be worse in percentage terms at both extremes of the power range.

Losses in diffs can be quite spectacularly large (10%), and again tend to be least at mid load."