If you draw precedents for a result for a collection, then rather than examining the precedent results of a derived results definition, it will first expand to show you the same result for the actual elements of the collection.
For technical reasons, if the collection is actually a collection of collections, then what you will always get is the actual members of the innermost collections.