I am not an expert in dynamometers, but there is definitely such a thing as one instrument being more accurate than another in its ability to measure torque output, and it is not such a gray matter as you seem to present it in your post.
Moreover, my understanding is that a dynapack chassis dyno has the ability to vary the load in real-time during the test to the extent that it can directly measure the maximum possible torque output of an engine at a specific rpm at steady-state. Note that a steady-state measurement at the hubs or the wheels would not take into account the rotational inertia effects associated with the drivetrain or the wheels, and will be the closest indication of output at the engine shaft. (You said above that the most accurate way to test motor gains is with an engine dyno). It is also my understanding that a dynojet dyno cannot do that as it cannot vary the load and simply infers the torque output based on acceleration time under a fixed load. Again, one is a more precise direct measurement method, and the other is based on inference. Correct me if I am misinformed.