This paper focus on query semantics on a graph database with constraints defined over a global schema and confidence scores assigned to the local data sources (which compose a distributed graph instance). Our query environment is customizable and user-friendly: it is settled on the basis of data source confidence and user's quality restrictions. These constraints are on queries not on sources. Inconsistent data is filtered, allowing the production of valid results even when data quality cannot be entirely ensured by sources. We elaborate on the process of validating answers to the queries against positive, negative and key constraints. Our validator can interact with divers lower-level query evaluation methods.