EIA insisting on plateau
If we average the two data sources in an attempt to determine the peak month:
Right now, May 2005 is the winner, but December 2005 and April 2006 are statistical ties. Minor revisions and changes to methodology could very easily change which is exactly the winner (indeed it was December 2005 for a while). But anyway, the oil industry doesn't seem to quite be able to tip it past the 85mbpd mark (on a combined agency basis) for the time being.
Finally, some folks expressed interest in just looking at crude plus condensate (ie real honest-to-God oil, rather than including natural gas liquids, biofuels, CTL, etc, etc). The EIA tracks that in Table 1.1, and here is the plateau in that:
As you can see, the basic plateau shape is just the same.
See here, for more background on these plateau posts if you haven't been following them.