They include a breakdown by date of onset as well, which isn’t exactly the method you suggest (but I read somewhere that recovery time is ~2 weeks for mild cases and ~3-6 weeks for severe).

Case fatality by date of onset:

[<Dec 31] 14.4% (n=105)

[Jan 1-10] 15.6% (n=653)

[Jan 11-20] 5.7% (n=5417)

[Jan 21-31] 1.9% (n=26468)

[>Feb 1] 0.8% (n=12030)

That does make the overall fatality rate look quite a bit worse than the naive calculation.

From the Diamond Princess we've got a pretty good idea about mortality. It's almost a conspiracy theory: let the disease run its course in a controlled environment to see how bad it is.

Diamond Princess isolated people, no? That is unlike what we'll see in the broader population.