Brian,
This another good point but there are a few things that will influence this with one of them is the engine calibration.
While actively participating in transmission spin loss testing, minimized negative torques are always used on the dyno during testing.
This is because the negative torque realized during engine braking is far less than the positive torque capability of the engine.
I mention the engine calibration because there are MAP calibrations used for vacuum control during deceleration to avoid excessive vacuum and engine oil 'pull-over'.
The decel control is not RPM based until the vehicle is near or at 0 MPH, until that occurs it is a 'target MAP' strategy or engine load strategy.
Additionally, these calibrations control how abruptly the vehicle crosses drive-line backlash and the related 'clunk' when you tip out of the throttle.
This can all greatly minimize the negative torque during decels, depending how they are calibrated, to minimize the thrust loading in the opposite direction.
On a side note:
The forces will be slightly different when comparing the 555/568 transmission families with their 520/523 counter parts.
If I recall correctly there is a minor pitch change between the low and high torque transmissions along with the obvious change in tooth count , their related taper and resulting tooth contact/engagement.
This will also be true/realized when combining gear sets to make a hybrid version of the above transmissions.