You're right, Matt. The acceptance of both accent and language effectively come down to the number of people using them - but that means the numbers of people within a community.
Two siblings who have a private and unique way of talking, which they both perfectly understand, are speaking a language with an accent. If only one of them does, then they are not.
The danger with the counter argument expressed above is that - because it is impossible to isolate and quantify separate, distinct accents - you can wind up dismissing entire continents as 'sloppy.'
Vis, a majority of speakers on the North American continent (and quite a few elsewhere) routinely pronounce the sound usually represented in 'English' as 't' by using the sound usually represented as 'd.' No-one objects (apart from himself above there), because there are so many people doing it.
But where do you draw a line between this huge population and the siblings with their private language?
Obviously you cannot.