I think it is necessary to render to HDR and use perceptual tone mapping as a post effect to simulate the dynamic range that the eye can perceive in the limited range that a regular display or a camera can handle.

As a simple example, this is how the eye would perceive a scene:

http://en.wikipedia.org/wiki/File:Old_saint_pauls_1.jpg

And this is how a camera would see it:

http://en.wikipedia.org/wiki/File:HDRI-Example.jpg

Of course, to get that dynamic range in an image, it has to be in the scene to begin with. Throw away pre-made light sets and build your own.