Before I worked at Google, I would frequently spend $10 - 15 a day in lunch costs. Add in dinner and snacks, and it could easily rack up to $40+. In pre tax terms that more like $55. Across an entire year that's close to $15k in benefit.
I work in NYC, so costs are inflated, and I realize this might not be true for people working in some less central location.
I'd rather take the free lunch than $1250.
And I'd LOVE to have on-site laundry and/or a shower - then I could bike much harder/further to get to work and not smell like it all day. Not having that means pretty strict pacing / length restrictions, which basically means I lose a large chunk of time that I could be using for exercise.
But to your first point: if all you're doing is reducing a part of a third of their caloric intake (one of three meals per day), you probably are not making a significant change in the person's health. It would be more practical to invest in company sports programs or fund things like bicycle purchases. The food is a psychological trick as well as keeping people planted at their desk throughout the day.
As to food: personally, lunch is my big meal, it comes out to almost half. Plus that's assuming all calories are equal, but meh. The main thing I would argue here is that investing in sports programs or bicycle purchases will probably impact a small fraction of the people that free lunch impacts. In addition, it'll have less effect on the people who could most benefit from better health - just having a soccer team or cheaper bikes isn't usually enough to convince overweight + sedentary to exercise more. It's pretty easy to get people to eat better if that's the easiest option.
This is also why I don't find a healthy lunch to contribute to greater overall health: people are different, and they usually need custom diet plans. Granted, lunch for everyone will impact the greatest number of people, but also in the smallest way. I submit that removing free sodas and adding tastier low-cal or zero-cal drinks would have an even greater effect on overall health than a free lunch, no matter how healthy the lunch is.
What we're talking about is trying to use convenience to covertly get people to be healthier, and that's really hard. If you really optimized you could make it so people biked or walked to work [vouchers on housing closer to work], place resources farther in the office so people have to walk around more often, a rewards system for taking breaks and walking around outside the campus [trackable by gadgets; also helps with vitamin d deficiency], allow people to bring home a healthy dinner, build a gym+shower into the property, and a rewards system for getting more sleep and regular exercise [trackable by gadgets]. The rewards could be tuned to the user's personal motivators.
That's all if you want to improve everyone's health. On the other hand, if you want people to think their job is cool because they get free lunch, you don't need to do those things.