[doh]

Well, not really. It's about how you store the data. For instance all graph data can be stored as hexastore [0] which can be stored literally in any database and you will get similar-ish performance out of them.

Some databases are optimized for certain workflows, so for instance they will store only "pointers" to certain data rather than copying them. In addition their clients obfuscate and abstract the storage layer from user.

But you can really store any data in any database, you just might have to do some extra pre/post-processing and multiple the storage requirements.

[0] http://karras.rutgers.edu/hexastore.pdf