My understanding is that at present, n1ql can only join on document IDs/keys, and cannot join across different document fields. I.e. there appears to be no current formulation to make a query work that would join across two properties in two different documents, neither of which is the actual document key.
Since one of the real values of n1ql is bringing a SQL-like language to document query (this is one of the things that lured me to it over alternatives such as mongo) - these kinds of joins in my opinion would be a killer feature for n1ql.
Is this planned on the product roadmap, or when might users expect to get it?