With a roll out you cannot measure the drive losses.
First of all you use the other side of a gear (other side of the tooth) which is not 100% representable, and secondly they don't know the total moving mass of the drivetrain, so they dont know the part that is loss and the part that is stored kinetic energy.
Imho it's an equation with 2 variables, so it's still an estimate. Maybe a better one (or worse one) than the competition but still an estimate.
If drivetrain loss wasn't effected by the power that goes through the drivetrain it would be (easily) measurable at the wheels, but it isn't.
Anyway, the engine power is not what gets you running, it's the wheel power, so it's best to focus on that.