We had a petri dish on Diamond Princess that might just give us truer numbers because of widespread testing, and the ability to follow afterward. 1.5% of those people have died, with more than 10% still sick (82 of the 701 who haven't died). That's clearly far lower than the 12% that have died in Italy, with studies showing real deaths are about twice that high. So clearly there's a lot of undetected disease out there.
I don't think in situations where the hospitals aren't overwhelmed, we'll see rates any higher than 1.5%, although the sheer number who remain sick for weeks is, to me, the biggest reason we need to limit those severe spikes in disease.