How to evaluate Database benchmarks ? What to consider given a specific example?

I'm trying to figure out which database to use for a project which is supposed to implement a temporal property graph model and I am looking into some benchmarks for that. I found some papers which provided some insights and results and I also found this benchmark from TigerGraph: https://www.tigergraph.com.cn/wp-content/uploads/2021/07/EN0302-GraphDatabase-Comparision-Benchmark-Report.pdf Does anyone have any idea why ArangoDB is performing so poorly here ? Especially in comparison to Neo4j ? Furthermore, any preferences regarding a NoSQL database which consistently needs to write data while answering mostly queries which result in large subtrees ? EDIT: Also - if someone has links to other benchmarks i'd welcome that.