It could also be that back in April not all patients that should have been hospitalised were indeed hospitalised (for lack of beds/doctors), presumably the threshold for "you should be admitted into the hospital" was higher in April compared to lets say July or August. As such, a lower "hospitalisation threshold" for July or August (compared to April) should have indeed caused a lower death-rate.

The criteria for being admitted was also stricter back then at least in NYC.