All models are wrong. Some models are useful - George Box famous modeler.
Everyone, especiall media, need to understand that all models are a range of probabilities. There was probability X that Hilary would get 51% of the Electoral college. There was probability Y that she would get 55% of the Electoral College. There was probability Z that she would get 45% of the Electoral College.
The same applies to the spread of a virus, and the deaths from that virus.
A model has variables: If people follow social distancing, the model predicts probability A. If people don’t social distance, the model predicts probabillity B. If we develop a vaccine quickly the model predicts probability C. etc.
Not only do the modelers not know exactly what the virus will do (as it is new and Chinese data is crap). The modelers do not know what the variables will do. The modelers do not know if we will social distance or not. The modelers do not know what orders politicians will give, nor whether those orders will be obeyed, nor if they will be timely. For most variables, the modelers know the probability no better than the role of the dice.
Then comes the political decisions. What level of probability is acceptable? What level of hospitalization and death is acceptable?
If a certain virus hits the GAY community hard due to the weakened immune systems of the GAY lifestyle, should we err more on the side of caution than if it hits some other demographic group harder? If the virus hits city dwellers who live close to each other more than rural voters who are inherently social distanced, then should the be a factor in whether to err on the side of caution?
A key task in analysis is to separate out (to the extent possible) the mathematical from the political. But of course, the sheer fact that we collect statistics on how it affects demographic group A but not demographic group B is political. Politics drives which facts are the convenient facts.