This question has no sense yet. The theory of patterns is only appearing. So far it is only a compact way to store template combinations for mapping purposes.Does your pattern have roles pointing at the yellow and the green instances of PossibleIndividual?
There is no RDF convention to store patterns, you can see an initial discussion for that at http://www.15926.org/viewtopic.php?f=3&t=160
iRING TIP Manager stores patterns in a relational database, my Editor is using Python dictionaries for this purpose.
Direct instantiation of patterns is theoretically envisioned by JORD pattern spec. There are no software to instantiate patterns today. iRING Tools users are probable thinking about it, but they do not need it yet. I can do it any moment RDF representation is agreed upon.
So far the only way to instantiate a pattern is to instantiate all templates included. My Editor does this, green and yellow roles filled by autogenerated URIs (like you see it for temporal roles in your templates).
Do not understand the question. It is semantically correct way to model engineering situation.If so, what do you gain with this denormalization? Can you fathom the combinatory explosion?
Proliferation of Scolem entities is no worse (and no better) then proliferation of temporal parts in exports of life-cycle data for your templates. There should be a business logic in the adaptor to ensure that only one Boiler individual is created for all patterns describing several different properties of this BoilerSystem.
Can not say any good thins about SPARQL. I just hope that at some point people will pay attention to my SearchLanguage (which already has pattern search) and understand that SPARQL should be encapsulated to obtain a higher-level language. From this moment technical aspects of SPARQL speed will be addressed independently of syntactic complexity. Or probable people will follow the road of http://www.optique-project.eu/ .