Goals against is trickier to judge and more to do with your own goalkeepers ability. The reason why people put value in xG is that most players over a long enough period of time perform pretty much close to their xG, be it Aguero, Lewandowski, Ronaldo, whoever. There are like 10-15% variances with some players like Messi and Martial, but it's not that massive and makes it a valuable marker. So assuming average finishing, then if you continuously create better chances then the opposition, that'll pay off and you'll get a hot run of finishing to take advantage of that. Of course finishing matters, it's what ultimately decides the results. xG is used to judge overall chance creation from a team, and it makes sense. Usually coaches develop systems to try to frequently create chances, while limit opposing chances. They don't design systems to create just 1 or 2 big chances and then tell their players they're shite if they don't take them. So I would say it's a good side just underperforming in terms of results if they have good xG but it isn't being represented in the real table. Yeah, it can last a full season, but eventually it will tell. Just like in 17/18 we didn't concede that many goals, but our defence actually wasn't very good and once De Gea stopped playing like a god we fell down to a distinctly mid table defence (like our defence probably was that year if De Gea wasn't in god mode).
Also a team being higher in the table compared to another team doesn't necessarily mean they are a better "team". Were we better then Liverpool in 18/19? We had more points, though they reached the CL final playing great stuff. Naturally if a team is on a clinical run of form that would lead to better results, so they'd be higher in the table. They did finish better, and that's the most important thing at the end of the day, but actual goals aren't good predictors for future performances and results. XG is a much better predictor for that, and statistically if a team continually creates good xG but just isn't finishing, eventually it should positively regress. The thing with football is it's always a small sample, so there is so much variability, but xG is the most accurate predictor.
Basically, see this:
http://11tegen11.net/2015/01/05/the-best-predictor-for-future-performance-is-expected-goals/
He explains it all. It's not the be all and end all, but generally a good sign of things working. As many managers have always said, a striker continually getting in chances is a good sign even if they've been on a bad finishing streak. You worry when they stop getting chances.