MAP Hypothesis and Consistent Learners

We showed that, in the given setting, every hypothesis consistent with $D$ is a MAP hypothesis.
Let a consistent learner be one that outputs a hypothesis with zero error over training examples.
Every consistent learner outputs a MAP hypothesis if we assume uniform prior over $H$ and deterministic noise-free examples.
For example, since Find-S outputs a consistent hypothesis, it outputs a MAP hypothesis, all without manipulating probabilities!