You mischaracterize one of MongoDB's main draws. It's not just indexing, but indexing over a schemaless dataset. Mongo isn't unique in providing this, but combined with other features it is compelling for certain scenarios.

That said, if Postgres matches/exceeds enough Mongo features, it could certainly outshine Mongo for those scenarios.

AFAIK, map-reduce is relatively expensive on Mongo... typically recommended only for background execution, nothing synchronous.