No you're missing the point. 10k or so is probably right but 1 million is way off. That number could be 50k (20% retention) or 100k (10% retention) or 200k (5%) but it definitely isn't a million.

I'm personally responsible for mining some thousands (maybe 10k or more I'm really not even sure of the number without doing some analysis) of those accounts with no user ever being associated with any of them (as well as creating probably a few dozen accounts, again with no users), and I'm not even remotely alone.

I actually think the high point in early 2018 probably represents a good fraction of the total users ever active, maybe 30-50%. For one thing a huge number joined during that period alone and for another the high price probably brought back a lot of the users who were here earlier and quit, if only to cash out some stake they may have left behind. So that puts the total at something like 100-150k and the retention at 7-10%.