I think the reason that it has been discounted is that the theory went that the first adult speakers developed a true pidgin (very little grammar, more words, gestures, and context) and then their children developed this rudimentary communciation tool into a full language over the course of a single generation. The idea is that before we learn our first language all of the possible 'settings' for language in our brain are in default mode and then as we absorb our first language these 'switch' into the correct setting for that language. So children creating a creole from a pidgin would use the default. This is already quite tenuous from a linguistic perspective (no evidence of defaults etc.) but also, it has been shown repeatedly that very few creole languages were actually developed in a single generation by children. I think from an academic perspective it would be very difficult to find strong evidence for a default grammar without some very morally questionable experiments. But I will point out that 'feral' children - who have zero linguistic input, don't develop grammars or the capacity for grammar, suggesting that there may not be a default.

Koko is an interesting one. She is a great example of an animal being taught communication but not language, the reason why we say not language (at least in its fullest form) is because she was unable to develop syntax or grammar. Just vocabulary.