Generally we judge the class of an object by its properties. i.e. If it walks like a duck and quacks it is probably a duck (apologies to ornithologists)
The sender and receiver often have a different classification systems for messages. e.g. Published specimen data is often(perhaps usually) received and treated as occurrence data.
I assume DwC is an generic way for marking up messages rather than a full blown ontology. The basisOfRecord and dcterms:type should therefore be thought of as hints on how to interpret a record rather than declarations of class membership. Class membership is decided by the receiver. An "unmissable special offer" to the sender may be a "useful fire lighter" to the receiver.
dcterms:type is used by the world and so should probably contain a generic term for systems/people who don't understand basisOfRecord.
basisOfRecord should contain anything that is well defined in natural language.

I suppose I am saying I don't understand the problem. Can some one summarise it?

Thanks,

Roger