This poses a serious problem to coverage as a method for evaluating the quality of the testing. For example, if we measure 55% coverage, it is not clear whether we have performed a really good test (that covered all the relevant functionality) or there is still a lot of relevant functionality left to be tested.