Paolo,
The structure of the model is inspired by the physical system, with the parameters determined by data fitting. I believe the people at Aalborg/NREL/Sandia ahve good faith in this torque loop model. However, Giorgio Bacelli is happy to provide further details on the origin of the model is you really need it. The direct feedthrough comes as a result of the fact that the force measured by the load cell is not exactly the force generated by the motor/gen but is the force exerted by the motor slider on the arm, while the input to the model is the desired/commanded force. The underlying issue is that the motor and slider/arm inertia are non-negligable above a certain frequency. If the direct feedthrough is causing you some problems (e.g. with algebraic loops, etc), you could put a z^{-1} (delay of 0.001s)………..