I may be wrong but I think this is the wrong way to look at it. They looked at 100% of the critical cases without any type of adjustment for age or anything and showed 82% reduction in death with statistical significance. Hitting statistical significance means that it wasn't a random event so if the entire trial had been made up of only critical patients then it wouldn't have mattered if there were 62 or 300, there still would be a benefit at day 14.
I think it's very appropriate to look at severe and critical patients separately for benefits (or detriments) like these. It would be crazy for the FDA to say "no, we can't give this patient in critical condition a drug that works for critical patients because it didn't show any benefit or detriment to the severe or moderate patients."