My mistake, it wasn't Harvard, it was Phillip Cook using the NCVS (National Crime Victimization Survey) Data. It's only been 8yr since I read the article, please forgive my indiscretions. In my defense Harvard is also mentioned.

https://www.npr.org/2018/04/13/602143823/how-often-do-people-use-guns-in-self-defense

Needless to say I disagree with the disqualification of defensive display categorically, but understand why it would be done for official estimates (though that would mean the official would be purposefully low, by the nature if many of them not having police reports or being one's word against another even if they exist there's no real reliable way to collect those statistics, so I think it's best to leave them out.) Mostly I think it should just be kept in mind that the official estimate is based off of incomplete data and is low, lest we end up with ridiculous estimates like 3,000,000.

And yet still, Harvard the NCVS data* (whoops lol) estimates defensive gun use as 40k higher than gun deaths, and that's with less than 20% of Americans carrying daily. To say it's so rare it's a myth is to say all gun death (60k), gun homicide (12k), accidental gun injuries (~1,500) and deaths (~500), are also myths due to rarity. And also the OG Black Lotus card at 1,100 printed.

The rate is far lower than the rate of gun accidents, or even gun homicides though.

Is still indeed not true, regardless of if I misremembered just who said that one detail from 8y ago.