Grades are how we try to measure knowledge, because knowledge is abstract and hard to measure. Obviously, the measurement of X runs the risk of being optimized as opposed to X - I see this in my job all the time, and we're all professionals.

Point is, it's not a problem with academia and it's not easy to solve. If we could just magically peer into people's minds and figure out what they know definitively then we'd solve a lot of world issues. Alas, we must settle for numbers and papers instead.