I suspect he was blood doping -- getting a transfusion of blood rich in red cells that he set aside six weeks or two months earlier, when he was using testosterone as a training aid. He probably just got the blood bottles mixed up and tapped the wrong one.
More on blood doping here:
http://www.slate.com/id/2107096/
The thesis of the article is that once a certain amount of doped testosterone is present in the body, it can take a long time for the levels to drop. How does blood doping account for the immediate drop in levels after the time of the positive?
According to Landis's website, he showed NORMAL levels of testosterone; what skewed the results was an unusually LOW reading for the E value. Thus the popular idea that he was loaded with 11 times the normal level of testosterone is false.