I realize that this question is addressed to taowave, but allow me to address it.
BS, with the appropriate vol param, returns good (not just decent, but objectively good) greeks. You can look at the options exchanges as markets for greeks, rather than individual options. As another poster (mrmuppet?) has pointed out, single name chains are neither deep nor liquid at the individual contract level, but are often pretty deep and liquid viewed as a market for greeeks. And, for any option position, a certain greek exposure is what you're really after.
You can, of course, with considerable effort, create an arb free price surface, fit some smooth function in log-strike and root-time to it, and derive relevant model-free greeks via finite differences or some similar numerical method. You will get, to within a few decimal places, the BS greeks. This has always amazed me, and is something I use (or programs written by me use) every day.
Similarly I use BS greeks to derive implied terminal distributions under the risk-neutral measure Q. Again, you can derive model-free RND's using quantile maximum-likelihood methods on those same smoothed price surfaces, but it is much more onerous, and you'll end up with the same exact (to a close approximation) terminal distributions. You can integrate these RND's times the payout function to recover all the options prices model free (without BS -- the prices match BS almost perfectly). This near-perfect match between BS and model free implied distros is also something that amazes me, and is also something that I use on a daily basis.
In fact an options pricing model can be thought of as a method to recover prices under the risk neutral measure. With a few tweaks (strike-expiry specific vols, mainly), BS does an excellent job of that.
