OK now I understand why I am not getting the published results.
Your testing method, while it shows what the servo is capable of performing in the lab, does not represent what it will do in the hands of the average modeler under operating conditions.
Its sort of like the way they used to test the engine of your car for HP. They took all of the accessories and fan off of it. Then the tuned it up to the teeth and ran it bare with not even a water pump on it. In the car it was not able to produce 1/3 the HP they claimed.