Puerto Rico report summary
I read it so you don't have to…
Basically says that they subtracted the reported deaths from their statistical estimate of what deaths should have been.
A large part of this estimate was further adjusted from average death rates because there was a huge migration after the storm, so they 'tweaked' the death rates to standardize the numbers.
Given the number is 3000 deaths, with a population of 3 million and a migration adjustment of 300,000… even a 1% error in the migration number could account for ALL the difference in the deaths.