Suppose there are three siblings age 31, 33, and 35. If the younger two of them die in an accident, would you then proclaim that the average age of death for the family is 32? Of course not, it can't be any less than 33--- and probably much larger.
Or suppose that you take a sample of the Winter 2012 Y Combinator class and observe that two of them closed shop as of 2013. Would you average those two samples and conclude that the average Y Combinator startup fails in less than one year?
While the Guardian article makes it clear they are talking only about failed products, if you want an expected product lifetime you have to include the "successful" ones too.