The behavior is quite consistent:
Under normal circumstances, up collective adds left pedal (increases TR pitch) automatically, and the pedals don't move at all. This mixing is done in the mechanical controls, via a mixing link. Down collective makes the TR go the other way. It does just what you would do, so that a liftoff to a hover takes virtually no yaw input. All Sikorsky helos after the S-58 (CH-34) have had this mixing..
IF the Tr is against a stop, so that no more TR pitch is available, then this automatic yaw input "bounces back" and is perceived in the cockpit as you describe:
If the TR is against the right stop (full right pedal) and the pilot pushes harder on the pedal to the right, he will simply back drive the collective and drive it up. Under this circumstance, no more TR pitch is being added, the TR is at its physical stops. The pilot is simply finding new combinations of collective/pedals that equal full TR pitch. It is somewhat rare, but certainly possible, to find this spot while taxiing, and also if one is in a right autorotative turn in descent.
The same thing happens with vastly up collective and left pedal, but the combination is very hard to find in normal operations. In that case, near the left pedal stop with near full up collective, adding left pedal drives the collective pitch downward.
If you ae still confused, I can draw a simple cartoon to show how the fixed TR setting creates a relationship between the yaw and collective.
BTW, if the TR control system should jam and freeze in place, this pedal/collective mixing will be seen in any flight regime. In a rare failure, due to a mode now fixed, a PHI pilot flew home and landed with stuck pedals and what he thought was stuck collective (but was not, the collective was free to achieve full range if he let the pedals move back and forth.)