It looks like the main difference is that SearchMonkey reads structured data, whereas Google infers structured data: even if the relevant data isn't marked on the page or in any particular format, Google will still find and surface it.
It's not so difficult to "infer" structured data from sources like Wikipedia. Dpreview is also an easy example. I don't see any proof the feature is actually about automatic inference, and not about hard-coded rules how to find structured data on some popular sites.