I have noticed that the position command and feedback values are always very similar regardless of the other commands, even if they are being sent wrong. If this is the case, does it even make sense to send velocity and torque/effort commands if the position command will overwrite them?
As long as these commands ‘agree’ they let you track a trajectory with much lower gains. If they 'don’t agree' - like if you send 'wrong' velocity or torque values as you described - the controller that ‘wins’ depends on how the gains are tuned, and frequently it is the position controller that wins.
We recommend playing around with adjusting the actuator gains in Scope or by setting gains with XML files in the Matlab API particularly raising and lowering the P Gain on position to see how it effects performance with and without commanding velocities. At lower position gains, you will better be able to see the effect that commanding velocities and torques have on tracking.