This relation is not formally defined by the GO, as I understand it. GO annotations are associations between a GO term (a type) and another class of biological entity e.g. a particular gene product (also a type, representing all instances of that gene product, rather than a specific individual), which makes things a bit tricky.

There is a discussion of the problem on the bioontology wiki where the author contrasts a more ontologically correct, but complex model, with a simpler approach that appears to be the one you describe for ONTO-Perl.