As the parallel with graphs seems to be a driving force in current discussions, a conceptual difference is the open world assumption of RDF in opposition of the closed world assumption that graph databases (well, databases in general) normally adopt.
In a web context, we should preserve the OWA as the default interpretation, but for use cases beyond web metadata, a lot of energy (and triples) can be saved if we can explicitly assert that a graph from a particular source is complete with respect to the properties it provides for the individuals that it contains. This would unleash advantages for reasoning and search and lead to more compact code.
As the parallel with graphs seems to be a driving force in current discussions, a conceptual difference is the open world assumption of RDF in opposition of the closed world assumption that graph databases (well, databases in general) normally adopt.
In a web context, we should preserve the OWA as the default interpretation, but for use cases beyond web metadata, a lot of energy (and triples) can be saved if we can explicitly assert that a graph from a particular source is complete with respect to the properties it provides for the individuals that it contains. This would unleash advantages for reasoning and search and lead to more compact code.