I have used Neo4j and Tigergraph. Neo4j's community version being a toy is something I agree with - haven't used the enterprise version, though. But its memory requirements do seem to be ridiculous. Tigergraph, OTOH, their free version could readily handle much much more data (and faster loading) than Neo4j.
But Tigergraph is UNIX only and built on C, not Java (probably an asset, TBH). Some UNIX sysadmin experience would be useful, but not mandatory, if you want to understand the innards of the software.
Cypher is perhaps a bit friendlier / easier to learn than GSQL. The necessary mental model for understanding GSQL takes a bit of time to comprehend.
I haven't looked at Neo4j in quite some time now, so I may be wrong here now - Tigergraph comes with some capabilities Neo4j doesn't like offering a built in REST functionality and easily embedded custom functions. When I thought about trying to do the equivalent with Neo4j, it would have required a microservice sort of proxy.
Neo4j has a lot of extra "stuff" in their ecosystem, like Bloom, but Tigergraph has been catching up in various ways, I think. Their documentation has gotten better since I first started using it and more add-ons etc.