@Neil: although low resolution PDB structures need to be refined, high-resolution structures are remarkably good (better than 1.5 Å). A reviewer for a paper I wrote recently made me look at high-resolution electron density, and with high-resolution structures, it's very clear where the heavy atoms are, and sometimes, you can even make out the hydrogen atoms. - Bosco Ho

I forget who (Peter Pulay?), said something along the lines of experimental science is nothing more than applying very simple minded theories :) - Deepak Singh

Bosco is right. Though the static nature of crystal structures is a major problem in my book, plus crystallization conditions are always artificial, which is another thing to be aware of - Deepak Singh

And isn't it sad that we are still using those same old force fields, or small improvements of those - Deepak Singh

@Neil: IMHO there is no such thing as "primary data" in crystallography, even the raw output from the detector is processed to "correct" for reflection angles, noise levels, sensitivity differences of different parts of the detector. - marcin