I would like to read this article - know the actual article/source? Good luck to anyone trying to estimate transmission/transaxle/gearbox losses with just one number.
Fortunately for once I can say here from lots of direct experience - speed, input torque, oil temperature, oil level, oil type, gear selection, age & break-in/wear-in, and even manufacturing variation & assembly variation etc., can affect efficiency &/or % loss measurements a lot. In other words, a single number for transmission/drivetrain loss makes little sense only under specific conditions unfortunately, and single % efficiencies w/o some tolerance are only "true" in controlled laboratories.
It was impossible to assign an automatic transmission a single efficiency number no matter how hard the program & marketing guys screamed for one from us because of the many different operation variables there are - was especially hard for auto transmissions because of the wildly varying operating conditions of the torque converter and the many various parasitic loss variations from all the different combinations of rotating clutchpacks & drums & planetary gears & seals that vary by gear selection, in addition to the variation of effic. across different rotating speeds, torques, oil temperature/viscosity, etc... For a manual trans car running at specific conditions, like WOT while trying to measure max Hp or Max torque, maybe a single % number could be used with some confidence but even then it should have some tolerance.
Anyway. Here's one time I'm not opinionating but instead barfing out info based on actual experience. Not trying to rain on any parade or squash discussion, but just suggesting: it's worth one's investigating/verifying further that using a single % effic number for a transmission doesn't hold much water and unfortunately might not be as useful a tool as one would hope.