This is so true. The hardest bugs and usually the ones that can get caught with a good code review are misunderstandings from the developer or business analysts not making the expected behavior clear.
Tests don't catch that. Unless the tests are written by someone other than the developer, which honestly I haven't seen.
As a contrived example if a developer interprets that they need to multiply instead of add, their method and tests might work 100% but they will give the wrong result. Even worse, in certain cases 2 * 2 = 2 + 2 so they will give the right result with the wrong calculation.