Right on. I have been in the position before where I needed to do some graph-like queries (where I had to traverse several levels of joins with the number of joins unknown a priori) but at the same time this was an application storing structured data. I found myself in a tight spot between doing some really clunky Postgres stuff or going the graph DB route. I would have loved to have something that can be used like a relational DB 95% of the time but can still handle that 5% of graph-like queries with reasonable performance and complexity.