[PeterisP]

By the way, in hiring you typically are ok with a reasonable number of false negatives, but you really want to eliminate false positives - so a strict filter is actually appropriate.

For example, if you're interviewing candidates with "true objective fit optimalities" of 31%, 52%, 84%, 91%, 94%, 96% - then your process must ensure that there is no way that the bad candidates get hired because they got lucky, even if it means that sometimes the top one gets thrown out by the filter. The delta between best and second best candidates is likely to be narrow (certainly much smaller than any margin of error of your interviewing/'measurement' process), and insignificant compared to the delta between a good candidate and a shiny-looking bad apple - except if your job market is starving and almost noone good applies.