What's interesting about graph is that it's really an ad hoc analytics use case ... it's not for operational / transactional.. this is what most people don't realize. If you have an at scale graph workload, like for example if you're facebook, you build your graph on top of an operational/transactional backend.. like in facebook's case heavily customized MySQL.

I bring this up because the ad hoc analytical use case for graph stores is so niche most engines haven't even seen enough demand to introduce it because you can always store graph relationships and offer retrievals in those engines to a limited depth which is typically sufficient for most operational / transactional use cases.