Note that the ship is travelling north in that graph, so new NMEA messages are often not represented by new reported positions for quite a while until the next "jump" occurs.

I tried .otherNavigation and .other Also, it is not snapping to roads, at all. It is just that when the ship moves slowly, it does not really update the reported position, insteads it stays around the current position for a while and later makes a big jump to a new area where it resides then. The behaviour seems to occur less often if the ship is travelling faster (e.g. downstream). I parsed the internal locationd logs for GNSS NMEA messages and resolved positions and related them. The NMEA reported raw positions are plotted in blue, the resolved positions are plotted as green circles with their radius being the reported accuracy, and there are faint orange arrows pointing from NMEA positions to the afterwards resolved positions which are then reported to the apps.

I finally was able to submit feedback FB20796539

We have a similar problem; we are tracking user's GPS location in order to play audios at certain POIs on a river cruise tour. The cruise ship is travelling very slow (3-5m/s) and calm. I saw from sysdiagnose logs that the NMEA GPS reception is very good, but many positions aren't used but instead sensor fusion reports old locations quite a long time with high accuracy, and then jumps to a new location, in a similar way than described here. This also affects other apps, including Apple Maps, and it renders our case unusable, unfortunately. I tried to submit the logs using Feedback Assistant, but the "Submit" button does nothing, unfortunately (nothing happens after accepting the privacy statement).