Because it can be misleading in the context of individual games.
A team can score a couple of really hard chances or deflected goals in the first minute, then sit back and allow the other team to have pot shots the rest of the game, that would then give the expected goals a false showing of the game.
It's a stat thats designed to be used over the course of a season NOT in individual games or a handful of games.
It's a metric that's most useful when considered in aggregate, yes. But that's not the same as saying giving up lots of shots after taking the lead in a game (which isn't even something United have done in any game I remember this season, really - more of a Tottenham thing) is a reason the aggregate figure can be distorted.
The numbers for an individual game aren't a simple "who should have won" indicator because of the need to account for game state, but they're still insightful - if a team goes 1-0 up early by scoring from a low-percentage chance or error, then sits back and lets the other team have 15 potshots for a total xG of, say, 1 or 1.5 and ends up winning 1-0, it's not wrong to say the other team should have scored once - they're more likely to score than not score if that same game with those same chances happens repeatedly. And a team that consistently sits back and invites pressure, even if it's comparatively low-quality chances, will give up goals and points over the season if it lets the opposition have enough of them.
The xGD numbers in that link the other person posted are an aggregate measure across half a season anyway. Although yeah, I don't think the difference (2 goals above XG for, 2 below XG against) is huge enough to suggest the numbers are saying anything about where the team should be in the table.