Right - and that's one of the reasons I don't feel comfortable with the idea of existential self-conformance. It introduces a new conformance for something that's really a language construct - not anything that exists in your model.
Personally, I'm leaning towards favouring the word "box" or "boxed" for existentials. It both describes what the thing does, and makes it clear when you should reach for it (boxing is a familiar concept in most programming languages and environments).