I agree.  But we do know with certainty that *some* have been lost, so the question becomes 'how many?'.  The Zombie Bitcoins article had a clever method for making an estimate, so that is why I used the 3 million figure.