Help with TrackObjects LAP - too many new objects

I am trying to track cells during a time lapse experiment, with an emphasis on correctly identifying cell divisions and cell deaths. I hope to use LAP, since it seems to be designed to identify such events accurately. I am assuming that a good 1st pass is important, but I am not getting great results so far. The problem is that objects that seem to be clearly trackable in subsequent frames are getting new object numbers in the 1st pass. For example, in the attached sample images, note the cluster of 4 cells on the right that have #'s 9,11,15,12 in Frames 1-2 and then #'s 29,31,33,34 in Frame 3. I have tried increasing the search radius from the default, and this has helped a bit, but not as much as I would like. Also, I don’t know whether I should be changing the search radius using the SD multiplier or the min/max thresholds or both.

As an alternative, the overlap and distance methods work fairly well, but I am concerned that I still may need to do too much manual review of splits and new/lost object calls.

Thanks very much for any advice you can provide,
4-28test_LAP.cp (8.7 KB) (121 KB) (5.49 MB)

Update - I had better results when I changed the LAP model from Random motion to Velocity. One detail that I did not understand from the help text is whether the velocity was estimated separately for each object, or whether a single velocity was used for the entire field. My cells are moving around each other in different directions, so each has its own trajectory but there is not a common directionality to the group. Since the Velocity model works better than the Random model, it appears that each object has its velocity computed independently.