It is even less than that. As you’ve discovered it doesn’t have the data. It packages the WN data as a module, loads the relations on demand and maintains a QLF (Quick Load Format) version of the files.
No. If you want to find exact whole words this is of course trivial. Else you can porter stem library to create an additional index for stemming, the metaphone library to create a sounds as dictionary or the library(isub) for loose matching. Finally the RDF library literal indexes can be used to create a prefix map. It all depends on what you want …
Note that SWI-Prolog will create all necessary additional indexes to make practically any of the lookups blindly fast.
The Porter stemmer is a “quick&dirty” stemmer and makes quite a few mistakes. Better ones exist nowadays, although I can’t help you find them (I haven’t worked on NLP for quite a while).