It supports both property graphs and semantic graphs seamlessly, plus a bunch of other powerful features Neo4j can't touch like virtual graphs (
http://docs.stardog.com/#_virtual_graphs), integrity constraints (
http://docs.stardog.com/#_validating_constraints), explanations (
http://docs.stardog.com/#_explaining_reasoning_results), versioning graphs with a Git-like model (
http://docs.stardog.com/#_versioning), user-defined rules (
http://docs.stardog.com/#_user_defined_rule_reasoning). And it supports all of TinkerPop3 including Gremlin.
It also has full REST support for non-JVM apps.
Stardog is a better graph database, but especially when integrating heterogenous data in the graph is key to success.