For example, in Michigan, their death counts were identical to what you see on other sites, up through April 3rd. Then on April 4th, for some reason their number of deaths is about 100 higher than the actual number. I wonder if it's a data entry error. As a whole, their data showed an increase in 600 deaths from April 3rd to April 4th, and since the 4th was the last day in their projections, it's pushing their projected peak to be very high now.
Next time they update their numbers, the projections will come way down.
The original projections (of deaths) were very close to the actual. It was only after their first update that it started to get a little weird.