Simrad AP25 NAV mode strangeness

I have a SImrad AP25 autopilot connected via N2K to a computer running MaxSea TZ. MSTZ is configured to output navigation data via a Maretron USB100 (which looks to MSTZ like a 0183 device).

Nav data seems to be getting to the AP25 correctly: when you press the NAV/WIND button, the waypoint data (well, some of it) is shown on the AP25 screen, and pressing NAV again hands control of the rudder to the AP25.

However, after running for a few minutes a problem shows up. The AP's XTE display will suddenly jump from (say) 100 yards right to 100 yards left, causing the AP to rapidly turn the boat to a new heading. A few seconds later it goes back. This happens repeatedly until I give up and switch back to AUTO mode and drive the XTE myself manually.

I'm trying to figure out what is going on here. One clue is that when I turn off XTE output in MSTZ, the AP consistently gets the heading wrong. This suggests that the AP is calculating an (erroneous) internal XTE which it uses when it doesn't get XTE from the nav source, and that MSTZ's XTE output is dropping out from time to time. When that happens, the AP's erroneous internal XTE drives the boat off course until the external XTE data reappears.

If the above is correct, the question is why the AP can't calculate a correct XTE. One theory I have is that the heading data supplied to MSTZ and the AP is different (it is; I have multiple heading sensors that don't all coincide perfectly). If heading is used to calculate XTE, that could cause the problem. However, I would think that COG would be a more useful input than heading for the calculation of XTE, given the natural errors in heading associated with magnetic variation/deviation, etc.

The manual doesn't go into much detail about the way that the AP25 calculates XTE. Any input would be appreciated.

Thanks!

/afb