You could do two things:
1) Use a microflow for the selector as the source of the data. Retrieve the values in a list, loop through the list to create a second list of the same objects and only add the values if the value for the attribute is not already in the list (with a find action list operation)
2) Create a separate entity and perform the same type of action as under option 1 but then to create the distinct values in the new entity and use this entity as the source for the reference selector.
I would choose option 2 for performance reasons.
Hope this helps you further in implementing a solution.