Those people pass both tests -- referral and standard hiring. To me the question hinges on whether people refer people who might fail standard screening, or if they're just cherry picking in ways your analysis "discovers."

The difference to employers might be moot I suppose, but if you want to substitute referral for standard hiring screens you kinda need to get at something like this to know if referrals are contributing any new information or just boosting hit ratios on existing tests.