You’re missing the point. We know for sure that in Lombardy the hospitalisation rate was higher than 1 in 1000 people infected.
We do not know the Lombardy hospitalisation rate because we do not have an accurate count of how many people have been infected in Lombardy.
To get such a count we would need 100% serological testing which has not been done. We know how many were infected at the time of testing but the whole point of the Oxford model is that we do not know how many have been infected, not shown symptoms, and recovered.
Once again, the core point of the Oxford model was to emphasise the need for serological testing.
An acquaintance's partner suspects having got it, because after two days (two days, not weeks) of very mild fever (~37.5C) he was hit by anosmia. An ex-coworker also mentioned "a horribly strong fever" with respiratory difficulties which lasted just a few days. Yes, anecdata, but shows that you can easily miss a large part of the infected cases if you only test those hospitalized.