Looking at the table above 4 results followed the XG prediction and 6 didn't.
I do think XG is a reasonable reflection overall though. The problem at the moment is that there aren't enough data points. One chance from the middle of the goal (say) 10 yards out can be very different to another from exactly the same place (I.e. is the keeper still on his line or right on top of the player, is the attacker in the clear one on one or are there defenders between him and the keeper, does the player have the ball under control, are they hitting it with their preferred foot, is it hit first time, is a defender putting pressure on, etc, etc). Over the thousands of chances from that position the XG might be (say) 0.6 but a team could have four shots from that position in the game and be given an XG of 2.4 as a result, ehenbin actual fact all four of the chances were 0.1 type chances due to one or more of the sorts of factors I mentioned above and the team was most likely to score zero goals.
To build on what Scaramanga said, see Michael Caley's xG map from this match:
You can see that there are chances from very similar positions that have very different xG - which shows that a number of other things are taken into account. (I'm not saying that means the model is perfect, but personally I think it's pretty decent and produces pretty meaningful stats overall).