Actually most studies agree with my experience. You can cherry pick some that don't but ultimately if people have the perception that X dog makes a good fighting dog, they train them to become a fighting dog therefore skewing the stats. It doesn't matter what dog you start with, you can make a golden retriever a viscous killing machine if you're abusive or encourage violent behavior, which is also abuse.
Well, sure, but if your goal was to create a vicious killing machine, would you be indifferent to the breed of the dog? Pits aren't good fighting jobs simply because people perceive them to be.