Hi Colette, saw your article linked from Adam’s newsletter - nice work.
I think you might be overlooking de-duplication (or lack of) using this method. Given that the link passes no time stamp, how do you resolve a scenario where multiple sources drive click(s) before the eventual install?
Seems that FB would get credit for every install they claim (even if other sources touched the user), which might explain why your numbers line up so well. Problem is, you might be double crediting.