xGF% is the division of xgf and xga. To answer your question, yes more shots against would raise the xGA and since Buffalo is not shooting as much you get that discrepancy. The way expected goals work, is each spot on the ice has a sh% more or less. So what is the average sh% from this spot in the ice over the entire league. This, as has been debated here, doesn't take individual skill into account so Thompson shooting from 10ft out is the same as Benson in the metrics. It helps to normalize and adjust for shots. The better models look at shot type, speed, and they are starting to examine pre-puck movement.
xGF is basically better corsi because it doesn't treat all shots the same. Teams that are -xGF% (so they are below 50%) are the teams that don't make the playoffs and lose more games. That is the general correlation and takeaway.