I am using the provided entities_names.json file from gnn/llm/README.md, but it seems to be incomplete. A lot of entities that are in the subgraphs in the training, val, and test sets do not have corresponding entries in the json. What should we do if we retrieve one of these entities?