Haven't read the paper, but it seems you are running the queries one by one and measuring the latency of each query. Please consider testing with concurrent queries.
TigerGraph always uses multiple threads to execute the query, so when running queries one by one, it gets an advantage by leveraging multiple cores. However, this also means it will scale worse than other DBs that executes queries single-threaded. The fact that TigerGraph seems to perform better in BI workloads can also be explained: multi-threading is more effective for BI workloads. A concurrent query test would be more fair for other DBs.
Haven't read the paper, but it seems you are running the queries one by one and measuring the latency of each query. Please consider testing with concurrent queries. TigerGraph always uses multiple threads to execute the query, so when running queries one by one, it gets an advantage by leveraging multiple cores. However, this also means it will scale worse than other DBs that executes queries single-threaded. The fact that TigerGraph seems to perform better in BI workloads can also be explained: multi-threading is more effective for BI workloads. A concurrent query test would be more fair for other DBs.