Knowing the principles of algorithms and how different algorithms for the same problem compare to each other is one thing, and expecting you to fully flesh out the details of an algorithm on a whiteboard or a piece of paper under a timed exercise is a completely different thing.

Some apparently very simple algorithms may prove tricky to implement.(Have you ever tried implementing binary search? :)) Another point is that we can't expect someone who's seeing a problem for the first time to easily map it to commonly known problem that has one or more specific algorithms to solve it. We as developers all know that even in our daily job sometimes it might take us several hours(sometimes even when we think as teams and not only individuals) to come up with a solution for a certain problem; this's happening even with problems and domains that we're very accustomed to during our daily work. We shouldn't expect better with problems unknown to interviewees.