Then your death rate is skewed because these are the cases that are (presumably) hospitalized. Not counted in your recovered to dead equation are all those with mild enough illness they were not hospitalized or diagnosed. Adding them in. would lower the death rate further
Yes, we can’t include cases we don’t know about. The same is true for any disease. Most people with the flu never go to the hospital. This factor is less significant for viruses like COVID-19, as it is contagious and dangerous enough that once several people around a “mild” case are hospitalized, significant efforts are made to track down the source.
It’s too early to assign precise numbers. In general, mortality will be low where good care is available, and higher where it is not.