"why did Peak Oil not occur as projected". Easy answer: when one is predicting the peak of anything one uses the historic population of that commodity be it oil, NG, iron ore, etc. So obviously one focuses on those production trends. Hubbert used the historic US oil trends in from the early 1900's to the 50's. More obvious: one cannot project the future trend of oil production until you have a significant amount of previous production from that trend.
Folks made bad predictions because they were not projecting production of trends that had not yet derveloped...like the Eagle Ford. One needs to remember that a Hubbert style production peak has no geologic basis. It's strictly a projection of a fixed set of distributions. If new trends develop that are not part of the original stats population any projection will be incorrect.
The simple white/black marble problem. One can pull 1,000 marbles out of a bag and develop a pretty accurate prediction of how many white marbles will show in the next. 1,000 draws. But if one dumps a big pile of white marbles into the bag between draws that original stat ain't worth sh*t. LOL.
The shale trends in the US are the new white marbles added to the bag.