That wouldn't necessarily be true even if 100% of submissions were good, as you've completely discounted the cases where no submission is made. You can argue about whether the whole thing has any point or use, but a low % bad on submissions isn't on its own a sign that such a filter isn't useful.
Looking at the funnel holistically, I doubt anyone who didn't write the essay would make it through the "standardized aptitude and personality tests". And even so - if the filter is submitted vs. didn't submit - there's no need to actually waste time reading them...
(If that sounds absurd, it is and this funnel is broken as hell.)
