With that large of a volume, the peer reviews are most likely only getting a cursory glance at best. Besides how many of the research studies are for anything truly worthwhile?
Furthermore ‘publish or perish’ is a very bad paradigm for true science to operate under. In fact science was a much more highly esteemed field prior to government grants run amuck amuck amuck...
Of course I have no idea what all those 1.3 million scientific articles published in 2006 were about, or how many were even in English versus Chinese or some other language, how many were funded by universities, or private enterprises, government agencies, national defense, etc, how many involved studies of exotic flora and fauna, planets around stars in distant galaxies, or even obesity in children eating McDonalds Happy Meals. ;-)
Regardless, I highly doubt if the accuracy rate implied by these numbers -- 99.9% -- is anywhere close to the truth, since doubtless many of those 1.3 million articles involved taking a fresh look at previous studies, and finding they were flawed.
So how many of those 1.3 million articles from 2006 will truly stand the tests of time and further scientific scrutiny?
Whatever the number, that's how science is supposed to work, and we should hope that it is working.