Basic Bayesian modeling. You work from a prior probability, then update the model based on data coming in. The difference you cite is easily explained by increased distancing/shut down measures.
It's not just me saying it. Even Captain Lockdown is saying it...... “All of the projections, by the way, all those statisticians have been 100 percent wrong at this point,” said Cuomo. “And we’ve been following the models because that was the only blueprint, but they haven’t turned out to be correct. “Sort of like the old days when we tried to predict a storm by listening to the weather forecasts,” he added. https://nationalfile.com/cuomo-every-coronavirus-projection-has-been-100-wrong-at-this-point/