The problem with xG as far as I can see is that all it really is is a more nuanced form of the shots at goal stat. It takes into account where on the pitch the shot was taken from, and (I think - going from memory here) whether it was a header or not. Nothing else. It doesn't take into account the position of defenders, or whether the chance is a clear tap-in or an overhead kick in a crowded penalty area. So as an example from the Arsenal vs Man Utd game, when Lukaku half went round Leno, who pulled off a fantastic save, and then had a swing at the rebound while off balance and with two defenders on the line, xG would give both those chances the same number as they were from the same spot.
As a statistic it has it's uses, certainly, and it is better than just looking at shots on goal, but it's a trap. The self-aggrandising name and the fact that it gives something approaching a scoreline leads to people over-valuing it's importance. It's easy to look at that 1.5 - 2.5 xG scoreline in the Arsenal Man Utd game and see that as somehow the "true" scoreline, when it was obvious from watching the game that Arsenal were the better team on the day and very much deserving of the win.
It gets misinterpreted too. When Burnley were doing very well last season playing a very defensive, very organised style, people brought up their xG statistics to try and show that they were somehow lucky, rather than that they as a unit were dogged and habitually got into position to block shots, and didn't let opposition forwards have clear chances very often. They defended deep and in numbers, Tarkowski and Mee are quite old school defenders who throw themselves into the way of shots, and behind them was a very tidy and in-form goalkeeper in Pope. Their xG was used to "prove" that their run was entirely down to luck, but xG simply does not account for good defending, or at least this kind of last-ditch defending. A more interesting use of the stat might have been to look into why they were performing above xG, but it's just dismissed as luck.
I am not against xG by any means, but it is not some kind of all-encompassing unifying stat that eclipses all other ways of looking at the game, which is how it's disciples tend to present it. It basically tells us that shots from closer in and more central to the goal are more likely to go in. Duh. I'm sure it's more useful as a coaching tool, and Guardiola can use it to communicate to his players that working the ball into the box results in better chances than shooting on sight.
It's just a stat, an oddly fetishised stat. Stats can only ever tell us a part of the picture. In the case of xG being used in individual games it doesn't tell us very much at all.