Farrell & Kammen meta-analysis, Science (Jan 2006)
The Farrell & Kammen study calculated GHG emissions for ethanol at the higher end of the confidence interval (95%). However, they did not compute the GHG emissions for the products displaced by coproducts of ethanol using the same confidence interval.

Clearly, you cannot legitimately compute one emissions amount (for the corn cultivated) at based on a 95% confidence level while using a point estimate for the GHG emissions of the products replaced by the ethanol coproducts. I have corrected their calculation by including a value for the GHG coproduct credit which reflects the high end of the 95% confidence interval for the products replaced by the ethanol coproducts: GHG emmissions for ethanol

Once you include the GHG emissions estimate for coproduct credits at the high end of the confidence interval (95%)(as was done with the GHG emissions for corn to show the possible (with a 2.5% likelihood) high end emissions figure for ethanol), the total GHG emissions for ethanol production at the high end of the confidence interval computes to 92 grams of CO2 equivalents per MJ. This compares to Farrell & Kammen's incorrect calculation of 121 grams of CO2 equiv per MJ of ethanol produced.

This puts the GHG emissions of ethanol at the high end of a 95% Confidence Interval below the average GHG emissions for gasoline. OF course, lest we forget, the probability of the actual GHG emissions from ethanol production actually reaching 2 Standard Errors of the Estimate above the Mean Estimate (or exceeding it) is 2.5%. This would establish this eventuality as what point spreaders call a VERY long shot. In other words the smart money would be on the 77 grams of CO2 equivalents (the estimate from the "Ethanol Today" scenario) which has a much higher probability of occurring.