You need to start with the basics. You need rudder
feedback and fluxgate compass data. Does the compass change as you move the boat. Is it accurate at least to each of the 90 degrees? Does the rudder
feedback display properly. If either of these are bad then you'll get that result. The next idea would be the course computer (the expensive part). You can check the rudder feedback with a multimeter. The rudder feedback is like the volume control on a stereo. It sends out a reference voltage and gets back a reading. There is a calibration for center and the computer does the rest. You can measure the return voltage where it connects to the computer.
To test the compass unmount it from it's location and slowly rotate it and read the display on the control head
. The compass works in 4 quadrants of the compass so testing the transition between them is what you need to validate. If the compass passes mounted then it should be fine and not require testing unmounted. It is possible something steel
is now too close where it didn't used to be close. It happens a lot and folks are sure nothing changed. Someone put something away where it didn't used to be and thinks nothing of it.
If you can validate the reading on the control head then you know these two are not the problem. Pray for a bad rudder feedback - they are the cheapest. The compass is only a tad more. The control head would be next at about triple the compass.
I've been down this road twice on two different boats. You need to eliminate what is not a problem then zero in on the rest. With simple tools and thinking it through it is possible to troubleshoot this.