Your "example" of bad data is the only quoted data in that article and it doesn't even mean anything since the implicit assumption of the Gupta model is that we do not know how many asymptomatic cases of COVID exist (even WIRED concede this point). Testing in Italy is insufficient to tell us this. 1 in 1000 infections requiring hospitalisation could be a realistic number if a high percentage of infections are asymptomatic.
> we just won’t know the true proportion of people who have contracted the disease without showing any symptoms, but it is likely a much lower number than the Oxford study assumes.
Epidemiology is not done by "it is likely much lower than the study assumes" since that is pure guess work without the tests.
What the Oxford study offers is a strong argument that antibody testing is vitally important and nobody is doing it.
As I say this is a very long way from utter rubbish.