Well, I'm no behavioral biologist, but I figure I can construct "just so stories" with the best of them. So here goes....
Fear is a biological reaction, like pleasure or rage, and is designed to perform certain basic survival functions, like triggering avoidance of certain perennial dangers (snakes, heights) and assisting us in learning to avoid newly encountered threats (hot stoves, musclebound bullies). Like other limbic-system mechanisms that tie into higher congitive layers, though, it is prone to the creation of accidental superfluous associations. (A single adverse reaction to eating a particular food can cause a lifelong aversion to the taste of that food, even if it's normally safe and nutritious. And I'll just mention, without elaboration, the word, "fetish".) From an evolutionary point of view, this is perfectly reasonable; after all, it's far better for an individual to develop an unnecessary fear of, say, rabbits than to fail to develop a fear of tigers.
The lack of a correlation between overall fear level and overall safety level is perhaps more interesting, because it's even more strongly counterintuitive. Our anxiety levels do seem roughly to track our relative perceived danger levels; that is, we tend to get more fearful when we sense great danger. Shouldn't there also be, then, some correlation with absolute danger levels, which have in fact declined markedly over the centuries?
I can imagine a number of possible answers: