Well, that's related to the particular implementations, not theory. If PostgreSQL works faster and is less resource-intensive than MongoDB, it says nothing about their theoretical underpinnings. The whole point of the NoSQL movement is that one approach (relational databases with SQL) to work with data and its structure cannot be the best answer to all the possibilities in that space. Not because of particularities in the implementation (eg. speed), but because it gives you an easier time working with it; there's no mismatch.
I'd say that if you are working with diverse documents with a few or no relationships, a document oriented database is necessarily a better fit to the task at hand than a relational one. It's a tautology.
If you are needing many JOINS and your data is homogeneous, there's a mismatch. You probably should have chosen a relational database in the first place, as it better fits that model.
Agreed. It's hilarious to watch people who keep trying to setup this RDBMS versus NoSQL situation. As though every data model can be perfectly handled by a RDBMS if only I learnt enough SQL.
I'd say that if you are working with diverse documents with a few or no relationships, a document oriented database is necessarily a better fit to the task at hand than a relational one. It's a tautology.
If you are needing many JOINS and your data is homogeneous, there's a mismatch. You probably should have chosen a relational database in the first place, as it better fits that model.