1. Patriots offense (pretty bad) is worse than the Chiefs' defense (average)
2. Patriots defense (very good) is worse than the Chiefs' offense (great)
3. Special teams is not included in this graph. I don't have numbers on overall special teams, but I know the Patriots surrendered 3 kickoff returns for TDs and had either the worst or second worst punting unit in the NFL, so that cost them a fair amount as well.
4. There's luck/clutchiness involved in converting expected points to actual points, and then again luck/clutchiness in turning actual points to actual wins. The math says the points for/points against for the Patriots would normally lead to 9.0 wins in a 17 game season (they won 8) and the Chiefs would normally have 11.4 wins in a 17 game season (they won 14). So their expected wins are quite a bit closer, but the Chiefs were better at closing out wins, and the Pats blew some close games.

These numbers are scaled so that the best offense and the best defense are 1.0, but the EPA/play models do tend to show good offenses with higher EPA numbers than good defenses. But that may reflect the EPA model more than the actual truth of the connection between offense and defense and wins. The 49ers likely have the best defense this year, and the Chiefs or Bills likely had the best offense, and all three teams have 13 or 14 wins. So... both are good?