The vision of the Semantic Web is becoming a reality with billions of RDF triples being distributed over multiple queryable end-points (e.g. Linked Data). Although there has been a body of work on RDF triples persistent storage, it seems that the problem of providing an efficient, in terms of query performance and data redundancy, inference enabled approach is still open. In this work, we take benefit of recent papers proposing a vertically-partition approach implemented on a column-oriented relational database management system and extend it with a novel approach enabling to represent less relations, thus requiring less joins in practical queries. This extension uses the fact that properties are first-class citizens in the RDF model to propose a particularly efficient solution when inferences are performed on property hierarchies. Another contribution of this paper is to propose a set of semantic query rewriting rules to improve query performance by reasoning over the ontology schema of the RDF triples. We provide an experimental evaluation on synthetic databases which emphasizes the relevance of our two contributions in efficient RDF triples storage.