To give concise explanations for a conclusion obtained by reasoning over ontologies, justifications have been proposed as minimal subsets of an ontology that entail the given conclusion. Even though computing one justification can be done in polynomial time for tractable Description Logics such as EL+, computing all justifications is complicated and often challenging for real-world ontologies. In this paper, based on a graph representation of EL+-ontologies, we propose a new set of inference rules (called H-rules) and take advantage of them for providing a new method of computing all justifications for a given conclusion. The advantage of our setting is that most of the time, it reduces the number of inferences (generated by H-rules) required to derive a given conclusion. This accelerates the enumeration of justifications relying on these inferences. We validate our approach by running real-world ontology experiments. Our graph-based approach outperforms PULi, the state-of-the-art algorithm, in most of cases.