But they measure those using 'coast-down' method and I think that is dubious. During coast-down, the engine is disengaged from the drivetrain, so if for example, you did a run and measured coast-down at say 40Hp; then bolted a supercharger on, did a run and measured coast-down again it would still be 40hp, so the transmission losses would be the same.
That cannot be correct?
Yes it is, its what you say, transmission losses, so the losses AFTER you "quit" the engine compartment. It does not need to count losses inside the engine itself.