It depends truly on whether you want data integrity and ACID features of a Relational Database. If a Relational Database is built correctly using the Relational Model and E.F Codd's rules, you will never have a problem with having duplicate data, inconsistencies, and other maladies.(This is assuming you use a RDBMS that is worth its salt, like oracle or SAP ASE)
However, you also have the option of MongoDB, which as you pointed out, is very flexible. However, through my experience, you will have to do a lot more manual work ensuring data accuracy and integrity.
However, certain things are easier in it, and it is by no means not successful. I use Mongo as a data back end for simulation servers I run, and it performs beautifully. Where mongo truly exceeds is with its atomic documents, and that's where Mongo pulled ahead of other NoSQL systems like CouchDB.
What it truly comes down to is what kind of data you are storing. If you are storing relational data, use a RDBMS. If it is more document based, use Mongo or a similar data storage engine. I do not like the idea of choosing a data storage engine by what is popular or what is new. Use what fits your data.
I hope this answers you question satisfactorily, if not please comment below.