I wish I could find the blog post again, but someone did the math and found that for their commute, driving actually emitted less CO2 than biking if the extra calories came from meat. I think it still works out positive if the extra calories come from vegetarian sources, but the delta is much smaller than I'd hope. Of course, there are lots of other benefits to not using vehicles, but it was interesting to think about.

Does this include extracting the gas, refining it, transporting it and burning it?

Yes. You have to be careful which comparison you are making.

The one I am making is "If a human could eat gasoline, there would be orders of magnitude less environmental damage to do so compared to eating ordinary food"

Humans can't of course, but the guy claiming biking gets 3000 mpg equivalent is making the same comparison.

In truth if we're making atmospheric carbon loading comparisons between traveling by car of bicycle, the equivalent "miles per gallon" would be much closer to equal with cars and biking, especially if you fill yourself up with grain light, produce heavy organic diets.

Of course not. Those are externalized.