One reason for thinking that a measure of talent is poor might be that it is outperformed by other measures. There may not be genuinely good measures of talent. It does occur some sort of retrospective measure based on results is probably better than what the admissions office uses, but that is surely still not a perfect measure, and is also obviously not a practical option to replace what the admissions office uses (unless someone invents a time machine). Another reason to think a measure of talent is poor, though, and this is probably more applicable here, is that a measure may be considered suspect if there is reason to think it is really measuring something else entirely, perhaps because it correlates suspiciously strongly with factors regarded as independent of talent.