There are glaring flaws in these breed statistics, they do not take into account the population of each breed. In many human statistics the data are equated on a per capita basis. This is not done with breed statistics because breed populations are not determined. Additionally, there is no good way to deal with mixed breed dogs.
I'll bet you'll find the bite and fatality statistics indicate that the most dangerous breed of dog is "mixed breed" since it is the most populous "breed" of dog. Those compiling the data often add bite data of mixed breeds dogs to the breed they most resemble.