But efficiency of converting sunlight into energy isn't really what this argument is about, since we're more concerned with the CO2 output than the energy efficiency. The efficiency of human muscles factors in though, as it is only around 20%, compared to electric motors and batteries that are over 90% efficient. Regenerative braking is a big win for the electric car too, since it reduces the impact of the car's higher mass a lot. And when you look at the CO2 per calorie emitted by coal (and of course the US grid emissions are much better than this since we use many other better energy sources than coal), you get about 1.1g per calorie for coal compared to about 4.8g per calorie for fruit. So in a carbon emissions sense, food can also be much less efficient than electricity, even when that electricity comes from fossil fuels.
No single one of these factors is enough to tip the scales on its own, and of course the numbers can skew either way depending on what kind of food you eat, what kind of electricity you use, and many other factors. But I don't think the comparison is a totally crazy one.
Regenerative braking is fantastic as an idea, but it does not actually offset the mass in practice. Tesla's regen braking has all kinds of limits - 64% electric conversion efficiency right off the top, but it also caps the regen rate, it doesn't regen when it's too cold, and regen can't recapture a/c or air friction losses, which actually add up to the majority of a car's power consumption for highway driving. I backed that up with data nearby. Regenerative braking is a great example of how cherry-picking efficiency percentages from the components of a system doesn't really help us understand the overall system efficiency.
Anyway, I don't know what the end-to-end system efficiency of food is, and to be honest I'm skeptical of anyone who claims to, the few stats we have are already gross approximations that tend to lose their context and their error bars. Combining them is extra problematic. My really one and only point is that the scales are pre-tipped 20:1 because of the mass on the consumption end, so no matter what happens in between, in order for the scales to balance, the complete end-to-end efficiency of electric system plus electric car manufacturing combined has to be more than twenty times that of food + bike manufacturing. No matter how good electric is, the savings are very much hampered by automobile manufacturing and by power generation. So it's just not that likely that the complete system for e-cars is a factor of 20 more efficient. There are multiple parts of the process that are the same or worse.