The cones aren't a placebo, they actually alter driver behavior. It's important when challenging an idea not to exaggerate its flaws. No practical number of orange cones will physically protect cyclists in a collision, but the psychological impact of well-designed infrastructure can have substantially better outcomes, which is why some places are much safer for cyclists than others.
That said, not all bike infrastructure is equal. Some of the new "protected" lanes in Seattle, for instance, are quite dangerous and I prefer to remain in general traffic.
You're right that cyclists should ride as if everyone driving a car is trying to kill them. But one effective way to do that, if designed well, is to actually separate car drivers from cyclists. Even an artificial separation can save lives, if it's done right.
One more bone to pick: making cyclists feel safer isn't necessarily dangerous. One of the reasons cycling is unsafe is because so few people do it, and one of the reasons so few do it is because they fear for their safety. There are even people who suggest that helmets are in aggregate counterproductive to bike safety for this reason (and I think they have a point, but I wear a helmet for other obvious reasons). Better perceived safety leading to more cyclists on the road will almost certainly result in better real safety outcomes.