I've written on the same issue. If you provide any kind of framework at all, then by definition, the learning is not completely unassisted. The term "unassisted discovery" is intended to discredit constructivist learning environments, and isn't a good description of how these environments are actually constructed.
Look at Seymour Papert's Logo programming environment for kids, and at first it seems like it could be "unassisted discovery." However, the structure of the language provides scaffolding for discovery far above what would happen if kids were doodling on paper (not to bash doodling on paper as activity...). Doodling in Logo is an incredibly mathematically rich experience.
I have some research which supports discovery based learning, although it is not as exhaustive as your list, but it does come from a variety of different fields.