The license indexer does not know in advance all foss licenses.
As of now, when unknown licenses are discovered, it simply logs a warning about it and go on with its life.

It's then up to us to check our logs, update the model, and then reschedule the impacted contents for indexing.

We can improve on this by adding unknown licenses directly before doing the actual association between content and licenses (implementation wise, it could be done in the storage).

A similar behavior is already done for example for 'person'.