If you look closely at the first graph you'll see how Laherrere models each liquid individually. Noteworthy is the URR for crude less extra-heavy, 2000 Gb, the same number got by Deffeyes (know what method he uses?)

Hubbert's method is not supposed to work when you model several liquids at the same time or several discovery cycles for the same liquid, that's what Laherrere shows with the last graph.

I would advise you to read an older Laherrere's paper:

THE HUBBERT CURVE : ITS STRENGTHS AND WEAKNESSES

Homework: apply the Hubbert Method to each discovery cycle in the UK and comment the results.

Laherrere's linearization graph has this epitaph:

Annual growth versus cumulative production: difficult to extrapolate when multi-linear!

You're trying to put something in Laherrere's mouth that he didn't say. If you want to do so, do it in your blog and leave it out of TOD.

The purpose of using Hubbert Linearization (HL) is to give us an estimate of the area under the curve, ultimate recoverable reserves (Qt).  As I described before, it's a gross simplification, but you can loosely compare oil reserves to water in a bottle.  You can pour it out a low rate, all the way up to the maximum rate, without affecting the total volume of water in the bottle.

Khebab, using only pre-1970 Lower 48 production data, predicted post-1970 cumulative Lower 48 oil production using HL.  Actual production was 99% of what HL predicted, through 2004.

Khebab, using only pre-1984 Russian production data, predicted post-1984 cumulative Russian oil production using HL.  Actual production was 95% of what HL predicted, through 2004.  

Now, the production profiles were vastly different, i.e., different rates of pouring out the water, but the rates did not affect the volume of water in the bottles.  

This was not a case of curve fitting.  Khebab is a careful objective scientist, and at my request he simply used the historial data to predict a produciton profile, that could then be compared to the actual production.  

Deffeyes is predicting that we have used 50% of conventional crude + condensate recoverable resrves, and based on the above case histories, we have to assume that he is right, and all of the data that we are seeing right now suggest that he is probably right.  

The HL method does not apply to nonconventional sources of oil, GTL on the light end and CTL on the heavy end.  But all we are talking about doing by going to the end points is increasing our rate of extraction of fossil fuel resources, which are already being commercially exploited.  IMO, nonconvetional oil will simply slow the rate of decline of our total oil supply.

As I pointed out in January, the Russian HL plot is suggesting a severe production decline ahead.  I find it ironic, that just as we continue to get more evidence that Russia is going to have--as predicted by the HL method-- production problems, the HL method is described as being "completely worthless."

How severe can production declines be?  Consider Cantarell, the second largest producing field in the world.  Internal Pemex reports suggest a decline rate of perhaps more than 40% per year, which seems reasonable when one considers the fact that the remaining oil column of 825' is thining at the rate of about 300' per year.

Texas, the Lower 48, Total US, Russia and the North Sea have never equaled the production that they had in the vicinity of 50% of Qt.  To argue against the HL method is to argue in favor of what we have never seen in large producing regions, i.e., conventional production increases past the 50% mark, in excess of the production in the vicinity of the 50% mark.

You can argue all you want to about the volume of liquids from nonconventional sources, but that is not relevant to the HL method.

In the case of the US48 peaking circa 1970, MKH hit it on the nose largely because it was just one piece of the world's production.  As the US48 approached peak, there was plenty of other production and therefore no price increase that changes consumption patterns and financed R&D for alternatives and the search for non-conventional sources.  We just bought oil from other places and went merrily along, so the US48 was able to follow its own peak curve with no market interference.

When you model the entire world you drag in a host of mitigating factors, most notably the market response to rising prices as you approach and pass the peak.  And in this context "market response" includes the development and use of alternatives and the exploitation of non-conventional sources.  

This is why I've been preaching for a long time that the "bumpy plateau" scenario was by far the most likely outcome: geology + economics = a slightly shorter and slightly delayed peak, and one that's much flatter.  If things play out that way it's extremely good news, as that would give us a prolonged period of higher prices that would spur the transition away from oil consumption before the actual decline sets in.

Basically what Lahèrrere is saying is that the production mid-point will be several years after peak production because of technology and non conventional sources (the production curve will be skewed). I don't understand what is defining as extra-heavy oil (EH), I guess it's synfuel from tar sands + NGL. I'm having a hard time in believing that EH production alone will reach 40 mbpd in 2045. For instance, oil production from Canadian tar sands is planned to double in 2015 to reach only 2.0-2.5 mbpd. This small increase alone will require a huge investment in new infrastucture. I think he applied a HL fit on the EH production that gave him an URR of 1Tb for EH. The problem is that the EH production is very immature (<25% URR), in that case the uncertainty on the URR estimate is fairly large (about 10% to be correct within 10% of relative error).