By “most” I mean “the vast majority” - at least 95%. And I only say that because in any dataset involving the real world there will always be edge cases that don’t fit cleanly into one category or another. The fact is that if you compare the official COVID-19 death count to the “excess mortality” figures (the number of deaths this year minus the number expected based on history) there’s a shortfall. That suggests that if anything, COVID deaths have been undercounted.
