I have had some issues with the Raymarine Autopilot due to the fluxgate compass. The most recent one was due to a local magnetic anomaly, causing a large drift. Nothing can be done about that one other checking the charts and being aware of the issue.
When I first purchased the boat it had been out of the water for 2 years. The autopilot tended to lose the plot completely initially until I had used it for about 2 weeks. The problem was linked to the signal from the fluxgate. It could have been due to it being a bit damp, but I can't be sure, and it did settle down after the electronics had been used for a while and I had a dehumidifier on the boat to dry everything out. The autopilot seems to be driven by the fluxgate compass and not the GPS informed COG heading, with some kind of alarm from the chart plotter if the difference is too great. Can anyone confirm if this is the case?