Why is that? MongoDB is very cool. It might be a little memory intensive for me but everything else is great.
There's nothing it's particularly good at, and several things it's outright bad at - data integrity being an important one. MongoDB has an
extremely poor track record in data integrity - to the point where until fairly recently, it would quietly throw away any data it couldn't store, which would be anything above some 3GB on a 32-bits system. Emphasis on 'silently' - it wouldn't even give you an error, just pretend that the write succeeded.
Other problems include it being non-relational (almost all data is actually relational), not having schemas (basically all data has schemas), hard to maintain, and performance not being very good.
Here's an article on the reasons why a company moved from MongoDB to PostgreSQL,
here's an article discussing data integrity issues during replication,
here's an article about performance, and
here is how you can store schemaless JSON in PostgreSQL if you absolutely need to (although you should use JSONB rather than JSON).
TL;DR MongoDB is primarily hype, it doesn't have any actual redeeming features. It offers 'fake' simplicity by not offering you a schema or meaningful relational constructs - things you will basically always need anyway, and will just end up reimplementing poorly. There is absolutely no evidence that MongoDB 'outperforms' anything in a real-world setting besides CouchDB.