I mean, it seems strange to reject implications of a successful model if those implications aren't observable. Eg, I believe that stuff exists outside our light cone, because it would be unparsimonious for there to be an extra rule saying that stuff that I can't access doesn't exists.