In any drivetrain component with meshing gearsets, heat generated by contact friction between the gears is a significant contributor to drivetrain loss. This is true during steady-state driving, but is far more of an issue when the throttle is mashed to the floor and the resulting thrust force and angular acceleration builds up in these drivetrain components. The heat generated by this dynamic friction is absorbed by the transmission and differential fluid as well as radiated to the atmosphere through the transmission and differential housing(s), and in some cases, via a heat exchanger or oil cooler. This absorbed and radiated heat is literally the conversion of engine torque into thermal energy because you can't technically "lose" power, but can only convert it into other things (some of our favorites being forward motion and tire smoke).
It's also worth noting that the more powerful you make your engine, the greater the thrust force and angular acceleration it's able to exert on the drivetrain, generating even more friction and heat in the process. But because both steady-state and dynamic friction vary depending on engine speed, engine load and the efficiency of the engine and drivetrain's design (how well they limit friction and the associated thermal conversion of torque to heat), there's no way to apply a universal percent loss to it. Nor is it possible to apply a fixed drivetrain loss figure to your car (say 60 whp from my RevUp G35 example), because as you modify the engine and increase its output its ability to generate thrust force and angular acceleration also increases (though not in a linear fashion).
In the end, there's no easy way to estimate the drivetrain loss your vehicle experiences on the road or even on the dyno. Coast-down tests are sometimes used on a dyno to attempt to measure frictional losses, but because this test is not dynamic (meaning they're not done while accelerating, but rather while coasting to a stop with the direct drive gear engaged but the clutch depressed so that the engine and transmission aren't linked) it really only captures steady-state drivetrain losses as well as rolling resistance. So rather than attempting to convert your vehicle's dyno-measured wheel horsepower to a SAE net horsepower figure using a percentage or a fixed horsepower value, you're far better off accepting the fact that these two types of horsepower measurements aren't easily correlated and forego any attempt at doing so.