FWIW.
Rotating mass has a least as much effect as friction.
The pump is a big factor in horsepower loss. More than you think.
So is the helical cut on the gears.
While it may be a very small factor heat effects of viscosity on fluid vs. friction on helical gears has not shown any real effects in racing automatics.
If I had about 3 hours, and it wouldn't waste the bandwidth, I'd go through all the work I did on reducing horsepower loss a few years ago. I spent nearly ten years working on automatics in record holding class cars.
Here are a few things to think about: In a non transbrake equipped PowerGlide for a class car (like a small block Super Stock car), I ran a low volume pump, a roller bearing between the pump and direct drum (the rest of a PowerGlide is already roller), an aluminum direct drum, an aluminum direct hub, straight cut planetaries, an aluminum planetary ring gear, four direct clutches, two reverse clutches, and a low pressure valve body. Eventually, I rifle drilled the output shaft.
We even had some transmissions that had small pieces of o-ring glued to the lugs of the steels to keep them off the clutches when not applied. Alto has been looking at making those for a year or so.
I built a 400 with straight cut gears and aluminum drums and reaction carrier, along with a low volume low pressure pump, a low pressure valve body (the valve body doesn't regulate pressure like a PowerGlide does, but had to be modified to work with lower pressures), along with a custom set of roller thrust bearings, and it was nearly as quick as a basic Glide with an aluminum direct drum. It was nearly as light too. It was within a couple hundredths in an A-B-A test in a 396 powered Super Stock Camaro.
Years ago, one of the advantages of a 200 was the fact that the stamped drums were lighter than the cast drums of other three speed boxes. With the advent of aluminum drums, that advantage was lost. There are still plenty of 200s in fast class cars.
In many class cars, we found that the RPM drop when using the super low gear sets in PowerGlides was not a major factor, as it just put the car back "on the converter" using some of the torque multiplication of a good converter to make up for a slight lack of gear. I had a bunch of cars running 2.08 and 2.18 low gear sets. I even had a 1.98 low gear set in my Super Gas Camaro.
The big thing to remember is that when you get into all the trick stuff in transmissions is that like anything else, nothing is free. You can spend $1K to get 0.05 ET reduction. You can also reduce the time between overhauls by a factor of 10 to make the same gain. A lot of that stuff is great for class cars under severe restrictions due to rules, where a couple hundredths is a huge advantage, and can give you the record. But in the case of fast street cars and drag cars running under open rules, most of it isn't worth it. Beyond getting the right box and the right converter, you won't find huge gains anywhere. But you can spend plenty of money and create plenty of work.