The difference is in the bar for forcing companies to set aside parts of their screening process. To get them to do that - as we needed to under our old model, because we didn't have a notion of "recommended but without the requirement to skip tech screens" - we needed to be making very strong recommendations.
I'm a little bemused at the notion of evaluating someone "strong tech skills, but can't pass an interview" and determining that the appropriate recommendation is "talk to this guy, but -- unlike with most of our candidates -- don't skip the tech screens". The recommendation seems like the opposite of the diagnosis.
How does the new ability to make that recommendation address the original problem?
Because getting some leads with less certainty is better than getting no leads with higher certainty in a tight market? Wasn't this addressed in the post?