I think this is a mischaracterization. The relational model which motivated relational DMBSs is based on predicate logic. Mappings to graphs are obvious, but are not the organizing principle. This was one of the strengths of the relational model, encouraging a more flexible view of the data than graph databases had previously offered. In a complex relational schema, you can discover and work with all kinds of implicit graphs that were not originally intended by the schema design.