Evolution AP not steering to waypoints without getting lost

[Seacom]

I'm helping a customer troubleshoot his Raymarine system and it seems to have an odd quirk. If we enable the autopilot to hold a course, it will do so tirelessly and very well for any duration of time. If we enable the autopilot to steer to a waypoint, even on the exact same heading, it does so flawlessly for the first few minutes, then suddenly veers off and starts doing a bit of a zigzag around the heading line before resuming it even tho ours hands are not touching anything at all. The MFD that is supplying the waypoint data is an E127, both units are fully updated and a proper dockside was completed, a seatrial was done and there was 4 degrees of deviation on the compass sensor.

The attached image shows a (left side) track going in a NW direction, this was us using the AP to hold a course. We then used the P70R to redirect our heading until we had fully turned around and began the (right side track) SW part of the trip. Once we had that line established, roughly 150 degrees true heading, I placed a waypoint a mile or so ahead of us on the same vector. I put the AP in standby, and reengaged it with a Goto Waypoint instead. It picked up its heading, followed that course for a good 3-5 minutes or more, then randomly began some zigzags before straightening itself out again and holding the completely wrong heading, at which point we stopped our trip and headed to dock. There was only the 2 of us onboard, and we were upstairs on the flybridge when it all happened.

While underway I checked the EV compass sensor, it is still showing a green light on the heading, it is installed correctly, and there is no metal objects of any kind within 3 feet of it, and that closest object is the AP processor itself.

If it matters, prior to upgrading to the E127, he had a C80 Classic, and when using it to steer routes and to waypoints via the same Evolution AP it never failed him once. When the E127 was installed, I did software updates all around including the AP, thru it. I also redid the dockside and compass calibration routines before putting it into active use. This zigzag problem was not a one off, it happened to him every single time he enabled a waypoint to steer too, and never once left it's heading if just holding a course without waypoints.