Flood simulations and actual flood risk

Looking at the maps in the article http://www.theatlanticcities.com/neighborhoods/2013/05/way-we-build-cities-making-them-flood/5590/  I found a report by the Chicago-based Center for Neighborhood Technology based on 177000 local flood insurance claims that found no correlation between damage and floodplains, as this graph shows: 


It is a completely counter-intuitive result: are flood plains not always a good indicator of flood risk?

The main problem is that the flood decisions that local governments are making today not only determine how much damage future floods will cause locally, but also how healthy the community’s river and coastal areas will be. Chicago and its surrounding cities, for example, have artificially created flood-prone places simply by paving over the region's natural ability to manage excess water.

Impervious surface creates risk over a much wider area, and it is impossible to say that just because you are locally above the 100 year flood, flooding is not a risk.