Various online calculators and formulae tell me that my whole daily commute will burn somewhere between 500 and 1000 kcal. The raw physical calculations of power output and so on calculations come out at about 500kCal, but that assumes no wind and steady speed rather than the stop/start of the in-town part of my commute. Maybe that's what the more opaque calculators are compensating for? Anyway I'll use a middling range of about 600-800 kcal. I will recover the energy used in this commute by eating, I'm not losing any weight.
Let's assume I make up these calories by eating roast beef sandwiches (which is a highly appealing prospect, and represents a mix of meat and non-meat products). These sandwiches will contain 50g of bread and 50g of meat, which is about two slices of bread and a sensible amount of meat to fill them. (I will suffer the lack of butter and horseradish sauce or mustard as a concession to convenient calculation).
Each sandwich will provide 113 kcal from the bread, and 94 kcal from the beef, for a total of 207 kcal. So, I'll need to eat three or four of these sandwiches to recover from my daily commute. Om nom, and indeed, nom.
However, because I live in a country where "[a]griculture is intensive, highly mechanized, and efficient by European standards", the food I eat will have had quite a lot of energy used in the primary production and supply chain.
This DEFRA paper has some interesting figures on quite how much energy is us ed to produce food.
Bread requires 11.7 MJ/kg to get to my plate, beef an impressively large 44 MJ/kg. So, my three or four sandwiches that I eat to give me energy to commute, which contained 150 or 200g of beef and 150 or 200g of bread, took 8.36 or 11.14 MJ to get to my plate.
So, that's 8.36 or 11.14 MJ for 24km, or to put it into a slightly more common unit for such things; 34.8 to 46.4 MJ/100km
Using a conversion factor of 9.8 kWh/l for diesel or 35.28 MJ/l, it's equivalent to a car driving .99 to 1.3 l/100km. That's about 4 times more energy efficient than a good car (or twice as efficient as an averagely loaded car), about the same as VW's impressive prototype, but still not free.
However, CO2 emissions from the food production are comperable to driving. Using the Global Warming Potential (GWP) of 16 kg CO2/kg beef and .98 kg CO2/kg bread given by DEFRA; that's equivalent to 106g to 141 g CO2 per km, somewhere between a fairly efficient car (like a Prius) and a middling one (VW Golf or Ford Focus).
OK, so you might argue that I was being deeply unfair by choosing beef as my example. Everybody knows that cows eat the planet. Eating a lot of vegetables is better, but still not free. If I filled up on rice; with 9.8MJ/kg embodied energy, and a GWP equivalent of 6.4 kg CO2/kg providing 350 kcal/100g, that'd give a much better 8.3 MJ/100km, but still 133g CO2/km. Potatoes, at 4.5MJ/kg and 304g CO2/kg would work out at 17.5 MJ/100km but only 11.8 g CO2/km. (I've simplified to a 700 kcal requirement here.)
Land - 700g of wheat flour makes 1kg bread - that takes about 933g of wheat. 1kg beef takes 7-10kg grain. My sandwiches take 1.2-2.2kg grain to produce. That requires 1.5 - 2.75 m^2 of land to grow at a yield of 8 t/ha.
According to this oil seed rape produces 1322.7 l/ha of biodiesel in the UK. So, the land for my sandwiches would produce 0.20 to 0.36 litres of biodiesel. Burnt in an efficient car, say 4l/100km, that'd only get me 5km to 9km rather than the 24km I got out of my sandwiches. (I'm assuming the energy inputs to grow the rape and wheat to high yield are similar, and thus cancel.) So, it looks like biodiesel still doesn't make much sense.
So far, I've ignored the embodied energy in the car, and that turns out to be critical. David Mackay quotes 30500 and 76000 kWh; or 110 to 274 GJ. another source claims 60GJ. Assuming that the car is used for 15 years and only for my commute (and I retain the same commute for 15 years), 225 working days/year, that's a total of 81,000km. Amortizing the embodied energy over the use gives an additional 74 to 338 MJ/100km (60 to 274 GJ) - huge compared with the 3240 litres fuel used over 15 years at 4 l/100km, or 31.8 GJ worth. (Admittedly, that is the high estimate since the energy is only amortized over the minimum mileage.)
I've also ignored the embodied energy in the fuel, or "Energy Return On Investment" EROI. Figures for this vary widely, and while there's lots of data on oil production, there seems to be less on end-user results. Best guesses I've seen are around 6:1 to 10:1 for petrol, which means total energy used by cars could be up to about 15% worse than the figure for basic fuel use. I will generously assume that the people doing the life cycle assessment for the food production took this into account in their calculations. Even so, it's just a tweak, it doesn't really change the basic comparison.
In summary, it turns out that as a cyclist, I really shouldn't feel smug that my miles aren't contributing to global warming. If I eat rice or beef emissions are just as bad as the same milage in a car, though potatoes are much better. The energy used in the commute is in favour of bikes, but not as much as you might initially suspect, possibly only twice as good as an efficient car being shared. What does come down heavily in favour of cycling is that bikes are much cheaper to build than cars.