I really am not sure what's happening here, but I wonder if he manual editing leaves objects with 'gaps' in their numbering which then confuses TrackObjects in some way.
You might be able to check this by examining spreadsheet output of object data from the 1st pipeline - do the object numbers have gaps?
One quick thing to try is ReassignObjectNumbers after the manual editing step. It's a bit of a hack but I wonder whether the Split option will have the effect of numbering your objects sequentially - I am really not sure. A more certain way to do the renumbering would be to re-identify the nuclei using IdentifyPrimary after the manual step, but that involves a bit more finagling to get the object identification right.
Another very quick thing to try is an alternate tracking method in your second pipeline, to see if you get the same problems.
If these ideas do not help, someone else will need to help brainstorm and troubleshoot!