Benchmarking isn't about production levels but are useful tools to understand how things perform. With variations on a given benchmark over several iterations it is entirely possible to paint a detailed picture of most likely scenarios, good or bad. That approach is what we should all strive to achieve.
TLDR - MongoDB is faster than any other NoSQL database in a benchmark published without any configuration. The results can't be reproduced let alone validated. However, in a benchmark published with all of the configuration and all of the results...