My boat is a 2019 Cutwater 28. I have had the similar auto pilot controller problem. The auto pilot controller seemed not able to register to the chart plotter consistently. After bunch of trial and error, I found if the auto pilot controller isn't seen from the chart plotter, I just need to power cycle the auto pilot controller.
What this tells me there is a bug in Garmin's NMEA 2000 bus handling. Like all buses, the NMEA bus needs a bus master. This bus master is likely to be the chart plotter in my situation. When a non-master device first gets onto a bus, it typically does a broadcast telling everybody on the bus of its existence. When the bus master receives this broadcast, it registers the device. In my case, it is the chart plotter registering the auto pilot controller after receiving its broadcast. However, during a power up cycle, if the bus master powers up later than its slave devices (i.e. the auto pilot controller), the chart plotter will miss the opportunity to register the auto pilot controller. It is 50-50 chance for the chart plotter to complete its powerup cycle ahead of the auto pilot. Hence, the 50-50 chance to not seeing the auto pilot controller on my boat.
I am looking at this from how a typical bus operates. Once the bus master receives the slave device registration, it replies with an acknowledgement message. From there, both of them talk. If the slave device does its initial broadcast and without receiving any acknowledgement from the master, it should keep broadcasting. In my C28 situation, the auto pilot probably does one time broadcast and that's it. That's why power cycle the auto pilot controller fixes this problem.
Long and complex explanation....
Frank