xG is like every other statistic that gets published about a game - it can be useful, but probably shouldn't be taken as fact.
The standard stats that we get (shots/on-target etc) rarely ever tell the whole story and all of them, xG included, need to be considered in the context of actually watching the game. For instance, the Brighton match:
https://understat.com/match/18371
Saints: 14 shots, 4 on target
Brighton: 7 shots, 4 on target
That would suggest that it was a reasonably balanced match to some degree, but anyone who actually watched it would know we got our arses handed to us and at no point did we ever look like we'd win the game.
Coincidentally, the xG on that match also fails to show things brilliantly as well.
Saints: 1.71 xG
Brighton: 1.01 xG
The Perraud O.G. is given an xG of 0.0, which seems to twist the numbers a little bit, and the JWP penalty and follow up are treated differently as they're two efforts, with xG's of 0.76 and 0.37 respectively (1.13 combined) - so discounting the penalty, the xG scores would be:
Saints: 0.58
Brighton: 1.01 xG
I'd also point out that Solly March's goal was spectacular, and rightly given an xG of just 0.04 - i.e. only 1 in 25 shots like that would hit the back of the net, which seems about right.
So taken in isolation on a specific match, they obviously don't always tell the complete story, just like all the other stats. But over the course of a season across all teams/shots/goals/saves etc, it probably spits out a reasonable estimate/average of how good a striker has performed. We know Che misses a lot of chances that he should score and his xG reflects that; we know Kane + Haaland score a lot of their chances, and again their xG reflects that.
tl;dr - it's not an exact science, it _can_ be useful, but it's not total bollocks