We sample a small portion of the frequently cited Movie graph, download this as json, then convert it to our format.
match (n)-[r]->(m) where rand()<0.001 return *
Their graph has 31,157 nodes and 170,499 relations. We expect ours to be proportionally smaller.
Our fully merged composite ~0.3 percent. graph 
Our small sample is 0.1 percent. graph
Our big sample is 0.2 percent. graph
http://ward.dojo.fed.wiki/assets/pages/mock-graph-data/schema.html HEIGHT 200
We fixed and converted neo4j json. Our sampling method made for many duplicate nodes which we have subsequently composed individually and then composed both big and small. github
pages/neo4j-movie-graph
Neo4j goes to some length to allow us to try the live database with interesting datasets. Login here with user, password, and database "recommendations". neo4jlabs