Generally speaking, apart from a (SCORM) repository of learning objects you also need a sequencing tool that maps the learning objects to different courses or curricula. So, while the learning objects and their metadata are stored in a repository, this repository is not to be used as a course delivery system that by its content structure reflects the courses you want to run on your website. That would mean you'd have to duplicate content objects that are used in different courses or other contexts, as you say.
I don't know what your solution is, but in the context of OpenACS/.LRN Ernie Ghiglione and friends have nearly finished developing a SCORM-compliant learning object repository and the idea has been to develop the Curriculum module into an IMS-compliant sequencing tool. This was nicely summarized in a thread a while back: https://openacs.org/forums/message-view?message_id=102631
In that thread Ola directly referred to the issue of using SCORM objects and other content in different contexts:
"As its name suggests, the Learning Object Repository is the place where course content should be stored, each individual object in its canonical place. It is not, however, the place where context-specific relations between content objects, as they appear in different course constellations, are defined. The LOR should store metadata about the content that is true regardless of its context. Curriculum, OTOH, defines course constellations by mapping activities, which have their special kind of metadata also defined by IMS/SCORM, to objects in the LOR. A map representing one course might be very similar to and might overlap a different map representing another course, kind of like using views in SQL... By doing this we don't duplicate learning objects that appear in different courses."
