I don’t see anything wrong with this approach.
You might have some basic type that can be described in concrete terms. Now, just because an object of this type might be further classified according to some subtype, it does not follow that all such subtypes are just as concrete; they may in turn require further concretization, as it were.
Real-world example:
Person — concrete (non-abstract)
Sibling: Person — abstract
Brother: Sibling — concrete
Sister: Sibling — concrete