No such thing as a dumb question.
WAR is calculated based on FIP, not xFIP; xFIP is a projection of what will likely happen in the future. People tend to think that FIP is an ERA estimator. It is, in the sense that it's a better predictor of future ERA than past ERA is, but that is not FIP's main purpose.
FIP's main purpose is evaluating what has already happened. Based on DIPS theory, or Defense Independent Pitching Statistics, there is a belief that pitchers can only control strikeouts, walks, and home runs. Once the ball is in play, the pitcher has little to do with the outcome. xFIP's purpose is not an improved ERA estimator, although once again it is also that, but rather a FIP estimator. It was found that pitchers generally tend to give up home runs on roughly 10% of the fly balls they allow. Some pitchers are slightly above and below this mark, so you may adjust accordingly, but the concept remains the same. This percentage stabilizes around 50 FBs. That concept of sample size is very important, and I'll throw it into the OP right away.
So xFIP multiplies the amount of fly balls a pitcher gives up by 10%, to project what number of home runs we can reasonably expect out of the pitcher. This is then substituted back into FIP to form xFIP.
However, since xFIP is predictive, it's not useful for WAR; WAR is a framework that shows what a player HAS done, not what a player WILL do. So, for this reason, FIP is used in WAR instead of xFIP.
The reason that Mark B's FIP has constantly been better than his xFIP is because he manages to work the corners and induce weak contact. Some pitchers can indeed do this: Buehrle, Rivera, etc. However, it takes a very long time to be able to identify these pitchers, and it should be assumed that they are getting "lucky" and not just beating FIP until they reach the level of a guy like Buehrle.
Hope I answered your question.