One possible issue is from an oversimplification. In actuality the beta parameters are going to be functions of time and space. To some extent quarantining can account for this, at least the temporal aspect, if you can model it by some plausible kinetics.
Another issue is in the data itself. There are believed to be many mild cases. For these, recovery=infected (up to a time lag) and if both go unreported, this will have the effect of making recovery seem slower than it actually is.
That stated, both posts are really nice.
