Thanks Ray, Kate.
To me, Spine Title is a type of title.... what sort of title is it, it's the title from the spine. Clearly, it could only be associated with an Instance, as a Work does not have a spine. Following the first discussion point, having bf:SpineTitle establishes a pattern of using URIs to identify things rather than strings -- in this case it is identifying the type of title.
Thus the last example (and the ontology) would lose the variantCharacteristic predicate and the title resource would be a bf:SpineTitle instead of a bf:VariantTitle.
Further work, in specific communities, could extend the base list of variants as required without worrying about the ambiguity of different communities coming up with the same string to use to name different semantics.
Otherwise, I think the proposal looks excellent!