Pete317 has beaten me to the punch with that bit about the dodgy graph. But I'd add another comment on that point. The two y scales both refer to CO2 but, as Pete said, do so in different terms - ppmv for atmospheric levels and gigatonnes for anthropogenic emissions. I can't see any reason not to express them both in the same units unless it's to assist you in getting across the image you wish to present - AKA how you spin it.
Here's a graph I did a while ago (sources Mauna Loa CO2 measurements by Keeling et al and anthropogenic emissions by Marland et al):
It's fairly typical of the sort of graph you see in AGW scare stories in the media in that the two trends are nicely lined up to show an apparent correlation. All I did differently was to convert ppmv into gigatonnes for atmospheric CO2, and as it happened Excel lined up the trendlines just so without me having to play with the scales. Which is what I did next:
Even leaving the supressed zero alone and just stretching the emssions scale from 10 to 30 gigatonnes presents a different picture. By the time you present the data honestly, i.e. no supressed zero, same units and same scale you end up with this:
By this stage I had to remove the trendline for emissions since it was hiding the X axis, but the X axis is so close to the trend anyway it doesn't make a lot of difference.
The data is all the same, but presented this way it wouldn't scare anybody, and IMO that's what a lot of this sort of thing is about. Why use graphs like the first one if you're sure you're right? Unless it's to scare the public into giving their unquestioning support (and their money

been there, which is at least part of the reason I'm such a sceptic now).
Edit: incidentally, does anyone know what proxy was used for the atmosphreic CO2 in that CDIAC graph on the page stevei linked to? I've got emissions esitmates going back to 1750 but I'm using the Mauna Loa atmospheric ones for CO2, which is why my graphs only go back to 1959. I've always wanted to make a bigger graph with the Marland data. I'm also interested to find out if, as Pete suggests, they've got two data series on one line there or if they've used the pre-1959 CO2 proxy for the latter half of the C20th as well.
Edit: actually it looks like an ice core (probably
Stiple Station in Antarctica judging from the dates) rather than a proxy, in which case it certainly looks like some of the Manua Loa or similar data has been tacked on there.
