This source strongly suggests that the original claim is not true. I don't understand exactly what the X axis is meant to be, but if we assume the 2012 number is meant to represent 2009 graduates, it seems incredibly unlikely that the median graduate has made no progress after 12 years if ~62% of them began making progress after 3 years.

(Can you come up with a story where the two numbers might be consistent? Probably, yeah. Thus the impracticality of disproving weird conditional statistics like this.)