They're world class health data scientists, of course they know what they are doing.
The only "issue" with their projection is that they are "officially" only being fed numbers of deaths that occur in hospitals. There is a very significant chunk that are not being accounted by this; this is coming to light now in countries such as Belgium and France. Of course, one can argue they are accounting for it, as their death projection range is very large (25k - 150k). As far as I know, data of deaths outside of hospitals does not exist at the moment for the USA.
Their resource usage projections are on point.
They have continoully updated their projections based on new models and data, its all in their notes and FAQ. Some here just wanted to discount it because it was was too 'optimistic'
While some are wrongfully discrediting them for the entirety of their analysis, you have to be nuanced in looking at the argument and whether it concerns resource usage/case peak and overall deaths. Their death projection range is very broad for a reason; for it to be completely accurate, they do need to be fed data about deaths not occurring in hospitals, which does not really exist at the moment. In that regard, it is too "optimistic," even though that's wrong to say since their range should account for the vast majority of situations anyway.