Fair enough, but it's also advantageous to make it non-obviously fake. Consider the recent stories about services faking traffic/users in order to gain real traffic/users (ebay, reddit, dating sites, etc).
You have a very fair point, but it's tangential to the one patio11 is making; creating testing data for development purposes (that's never going to be seen by users, short of an accident) and creating faked data for marketing/growth/etc purposes (which is solely for the purpose of presenting to users) are unrelated.
I guess I'm not following how they are unrelated. To really stress a system and find bugs, shouldn't you be putting it under realistic usage and load? In my mind using test data that closely matches real data is desirable for development purposes. I suppose I'm now actually disagreeing with the OP's statement -- that example was a specific circumstance where perhaps more care should've been taken, but for the general case I think it's actually desirable to use realistic data.
Plus I just think it'd be cool if you could use the same service/API for both use cases.