You're confusing preseason projections and simply runs scored vs. runs allowed.
It is absolutely true that regardless of the projections or advanced metrics, when you score and give up a certain amount of runs is not a repeatable skill. Over 162 games, run differential is a far more accurate guide for how good you are and how good you will be going forward than wins and losses.
The Braves were as good as their record indicated they were. The Nationals were a better team than their record indicated but still not as good as the Braves. And no, they should not have won 96 games. They should have won about 89-90. The Braves were about where they should have been, if not a win or two below it.
The Mariners should have won 77-78 games based on what they actually did on the field, not based on some projection system. Thinking that team is an 89-win team instead of an 80-or-fewer win team is how executives look stupid down the road and get fired.