Well, my initial general thought is that as energy surplus declines, we will borrow what we can to maintain it from the poor (rising inequality) and the environment (higher externalities). Certainly neither 'straight' EROI, nor conventional economics accounts for environmental externalities, though the market is starting to in a small (misguided) way with carbon credits.

The other thing off the top of my head is that the canadian numbers have to be quality adjusted - if the main energy input is now electricity, but it used to be something of lower quality, then quality adjusted EROI (in the primitive form you illustrate) could be declining.

Also, its possible that what your chart might show is the 'cannabilization' of the in situ resource that is 'free'. The recent rise in EROI since 2004 seems to coincide with the scaling of tar sands. If a good % of the tar sands themselves are used as energy to process different tar sands to an end product, I doubt the NEB is counting that as an energy input.

Also - remember that energy surplus is EROI x Scale - so if total energy is declining, even with higher EROI you have smaller societal energy surplus (not sure if thats the case here)

Finally, the above chart doesn't tell us much about fixed vs marginal energy investments. Its possible in some of those missing data points that very large energy expenditures were made in energy infrastructure, and that we now have higher EROI because we are predominantly using built infrastructure as opposed to energy intensive new structures, to harvest the oil/NG, etc. We don't know though (but I suppose could back into that...)