Quads etc.
What would be involved in having not triples, but quads or longer sequences, as the base target for expansion?
As an example to indicate why this could make sense, consider the need for hierarchy among ontology modules: With an RDF store that has named graphs, more quality controls could be possible. We generally need to ensure that a class is not redefined in an ontology further down from where it is introduced.
I realise that this would need discussion, but it might be an unfortunate constraint to limit templates to triples only.