That likely has to do with availability of tests, who's getting tested, where and when. With Sweden not attempting to stop the spread, I doubt they've done wide-spread population testing. They're likely limiting it to hospitals, so the fact it's flat its really what matters until they kick off their wide-spread serological study in the works.

[edit] I misread, disregard, but preserving for posterity.

There is random sampling being done. 2,5% had an ongoing infection in Stockholm in late Mars.[1]

Now, 11/100 blood givers in Stockholm who has not been sick in the last two weeks have antibodies according to researchers.

That gives you a naive mortality rate of 0,4% per infection, assuming the spread among bloodgivers is the same as the population in general and just dividing the death count by 11% of the population. It is a high eastimate as I guess many wont get traceable antobodies and mostly asymptotic people give blood.

[1] (Swedish) https://www.folkhalsomyndigheten.se/publicerat-material/publ...

Active vs. recovered is a metric related to the people who are tested and confirmed. It cannot be changed by testing fewer people, only by rapidly ramping the number of tests up or down.

Ah pardon, I misread. You're totally right.