Ballistic gelatin tests are used to PREDICT bullet behavior in actual shootings. The reason we know it to be a *relatively* valid test (given the limitations that we don't perform bullet testing on live humans) is because those results are compared to the results of many thousands of real-world shootings, from military to police to civilian, from all around the world. This feedback has allowed the testing to be refined to the point that it is today, and has been the impetus for adding the various "barrier" tests along with just shooting bare gelatin.
Were there to be a load that showed a trend in real-world shootings that didn't match the predicted behavior pretty closely, then perhaps there would be an issue (this WAS the case with bullets passing through barriers, hence those additional tests). So far, the tests as currently used continue to prove the best scientifically reproducable model for predicting real-world bullet performance.
In the real world, you never know exactly what tissue the bullet is going to encounter. Gunfights are extremely stressful, and those involved are usually moving (even if it is just the hands/arms moving in front of the chest to point a gun), and rarely present a squared, front view like target posters suggest. Think about what a bullet path would be when shooting at a guy driving a car, from the front, side, or rear. How about a guy lying prone? How much penetration will you need to reliably reach the vital organs in these situations? Remember that all of these are more likely than the bad guy standing at attention facing you while you pump bullets into his chest. Straight-on torso shots are the exception, not the rule.
-Troy