The activity types results list is ordered based on match probability. So if you never use an activity type, it will be at the bottom of the list almost always.

Note the number on the right side of the list - this is the normalised match probability score for each time, and the number used for the sort order.

Note that even types that appear to have a zero score might still have a score slightly above zero. This also contributes to the sorting, so that very bad matches are still sorted by best to worst.

The more corrections and confirmations you make, the better the models will understand your common activities and routes, and the more sensible the sort order in the results list will be.

So in effect, confirming and correcting timeline items is the feature you’re asking for. By confirming and correcting, you tell Arc which activity types to put at the top, and which to put at the bottom.