NET doesn't actually use scoring margin, but it does have an off/def efficiency component, so if you score more points per possession (and give up fewer) your NET improves. And especially early in the season, a single game can have a bigger effect on the efficiency metrics--should level out over time. KenPom is actually pretty similar in that respect, which is why the johnnies probably jumped a bunch there too (though I think KenPom actually uses old data early in the season to smooth those jumps--not sure whether that's actually better or not, probably not). both are "predictive" meaning they don't really evaluate a team's resume (although that is part of the metric) and shouldn't be used that way.