That was the reason why I currently have a separate RDF store for each actor, so that at least privacy is guaranteed in the event of a problem. I would then want to work with SPARQL endpoints per collection. How I then solve this SPARQL technically is still open, but I have a feeling that this will work 
1 Like
followup: the current solution seems to be putting these “private” details in a separate private graph (and then merging them as appropriate?) at the very least, WebID uses rdfs:seeAlso for this purpose, although this is maybe not the best way to do it. “see also” UsingSeeAlso - W3C Wiki for more details. (i would also point out that the assumption that the target will be an rdf/xml document because i think the semantic/ld web has mostly moved away from rdf/xml as the “default” or “only” repr. turtle and jsonld are certainly far more intelligible than rdf/xml.)