Choosing Between Graph Databases and Relational Databases for Connected Data?

Hey everyone, I'm working on a project where we need to model highly interconnected data, like social networks or complex dependency graphs. I've been looking into both graph databases (Neo4j, Amazon Neptune) and relational databases with graph extensions (like PostgreSQL with extensions or even just clever table design). What are your experiences with these approaches for situations where relationships are just as important as the entities themselves? Are there clear tipping points where one becomes significantly better than the other, especially concerning query performance for deep traversals or data modeling complexity? I'm particularly interested in real-world scenarios where you've had to make this choice and what factors ultimately led your decision.