It's a standard figure. Suspiciously round, basically unsupportable. It's really hard to measure what percentage of companies fail because so little of the data is public.
But the figure wasn't companies that didn't fail, it was companies that made it big. Surely that's an easier number to come up with, depending on your definition of "big".
tl;dr: 1 in 20 would be a better estimate than 1 in 10, but even this underestimates the skew.
Across all VC-funded startups which received first-round funding between 1985 and 2009, 55% were terminated at a loss, and only 6% returned more than 5 times the original investment. But this 6% group generated more than 50% of the gross return across all ventures.
At a "single large and successful" (but, for obvious reasons, anonymous) VC firm which invested about 1 billion over the last decade, about 5% of the total money invested generated a return of 10x or more; almost 60% of the money was invested into companies that terminated at loss.
You may also want to take a look at Figure 2 in Hall and Woodward (2008), which shows the distribution of exit values over 22,000 VC-backed startups founded between 1987 and 2008. About 5% of these startups had exit values of $50 million or more. (http://www.nber.org/papers/w14219)