Precisly wrong or roughly right

When attending the Product Owner Class, Mike Cohn gave the reasons for estimations being given using story points and in a relative scale.

The reason is that an estimation should be roughly right rather than exactly wrong. If you give an estimation which is 350 hours, the risk for this not being the exact answer is great. But to say something is about twice as big as X, that is a completely different ball game.

This also applies to other areas than estimation of development work. How about a map. The picture is of an add for google map in the NY sub. The problem is that the map shows the wrong location.