I would not produce a vendor or language specific file… just node and edge CSV and Parquet files. You could also do Cypher files, but node edge lists are the most standard, commonly used property graph format that work in EVERYTHING, including a Cypher DB. To support those, just include Cypher files that load the node, edge lists? Academics especially rely on node edge lists, why tie to one type of database when it’s more work? 
I get it’s about user convenience, but it’s a layer of obscurity if you’re not using Neo4j.
The actual config file looks okay, but for me I care less about multiple versions of a property graph export than ONE that you support. Know what I mean? I need that to build my library of network motifs for UBO, KYC, AML and CTF on top of them.
You may have read this post: https://blog.graphlet.ai/financial-crime-and-corruption-network-motifs-4cf2e8e10eb5
This draft is half done, it talks more about what I’ve got in mind and what I need: https://medium.com/@rjurney/9227825962c6
I hope that helps clear up my requirements some?