You spent that much time over several dev sprints/cycles/months/whatever. Meticulously designing the expected output and quality expectations(yet not sharing any of that in the spec supplied to the candidate). All that automated testing, manual code review checklist was designed after spending that many man hours worth effort to arrive at. You have the benefit of hindsight, and insights from the process that refined over that much time. Plus you have inputs from several candidates you failed your test.

Yet now the candidate has no luxury of that, and has to guess the assumptions required to not just produce all the features, but also the quality metrics the project will be measured against. All in the deadline given by you. And he is at day 1, and that's all the day they have.

This is really like expecting some one to fix a problem in a day, what the other person spent months solving.