What could be wrong with my EV-100 autopilot?

[dove]

There must be a simple answer to this. In short, the EV-100 AP (power version; no rudder reference) installed on my new boat works OK at low speed but has problems making course corrections to the right under resistance at speed. At the dock and at low speed, everything seems fine, and there is no difference in performance or feel right or left. But if there’s more resistance on the engine, as from running 25 mph, or even enough wind or current at low speed, the series of short pulses from the AP pump don’t make the intended corrections to starboard. The AP eventually believes it’s reached the set “rudder limit” of 25 degrees and stops trying, even though it has not moved the engine off dead center. This happens even though the AP is set to “performance mode,” intended to give the quickest corrections, and regardless of whether it’s in “auto” or “track” mode

Here are two interesting clues which might help. First, when the boat is veering off course to the left because the AP isn’t correcting to the right, it will quickly and fully correct if I just give the wheel a little tug to the right to help it. Second, regardless of speed or rough conditions, the AP will make any large maneuver to the right if I tell it to track to a course that is way off to the right, like 45 degrees. It has no problem maneuvering the boat around aggressively to the right when commanded. It’s the small corrections, or series of them, that seem to be ineffective against any resistance. With the AP disengaged, the boat’s helm performs normally as described in the background below.

My own observation is that the EV-100 software uses the pump in two different ways depending on the situation - either tiny, short pulses, including a series of them as for course maintenance, or long, continuous pump operation, as for acquiring a new course. It’s the first of these that doesn’t work for starboard corrections against resistance. What might be wrong? Is there anything else I should try? Could the pump have an issue, like a sticky valve?

More background: The EV-100 was professionally installed on my brand new 22' bay boat with a Yamaha F150 and Sea Star helm # HH 6541-3 and 1.7 cu. in ram # 5345. The AP seems to be set up properly with a hard over time of 14 seconds, which is the actual time measured driving the motor stop to stop in either direction with the AP pump. (I’ve also experimented with higher hard over times and other settings to no avail.) I have all the software updates except for the latest May 2016 updates which supposedly are not compatible with my Garmin chartplotter (though I would obtain the updates anyway if they fixed my problem, but that seems unlikely since no one else reports the same issue.) The boat’s helm by itself performs normally and smoothly in either direction, with 5.1 turns stop to stop (spec is 4.9 turns fully bled.) There is no feedback at the helm from either the engine or AP operation. At speed, it’s a bit easier to turn the wheel starboard, which is the normal effect of prop torque. If anything, that should help the AP correct to starboard. But as described above, it won’t stay on course at speed unless I watch and help it along.

If my pictures got attached, one shows the installation, and the other shows (a dockside simulation of) what the P70R looks like when it thinks the steering is fully to the right when it really hasn't moved.