Currently we store the contents of single embeddedables (not element collections) as a separate node. An alternative mapping strategy would be to store the properties of embeddedables within the embedding node itself, using a "dot notation" for the property names.

I'm not sure what of the two approaches would be more "natural" in Neo4j. Traversing to a separate node may not be as expensive as a join in RDMBS, still I'd assume that just getting a single node must be cheaper than getting the entity node, traverse to the embedded node and fetch it as well.

An option for allowing to configure the strategy to be used may be useful.