How MirrorSync Determines Who "Wins"

Hi. This may be a question with a very obvious answer.

I recently helped my client rollout MirrorSync on iPads. They have about 7 people on the field, each with an iPad. Ultimately, mirrorSync is being used because they need to take photos and store them in the database while out on the field. They are usually working on areas with little to no cellular data signal.

At first, they reported issues with errors during sync'ing. We were able to get around those errors by redistributing the same database file that we started with. However, recently, some of the iPads began having troubles not having the data that was expected on them. I did also noticed that some records (mainly, the "documents" table with a container field in it) that seemed to be missing a lot of records in the hub database.

To confirm, MirrorSync uses solely the number of modifications and the modification timestamp to determines who "wins" with updates, correct? How does it determine deletions?

A couple of the iPads have the container data, but not the hub.

Could I manually update the modification timestamps on the records on the iPad (spoke databases), run a sync, and use that to "update" the hub database? Or, should I gather the container and other data from the spoke databases, manually update the hub database, then run a sync?

Insights, suggestions would be very helpful. This is rather urgent.

Thanks for your help!

--Raf