The primary driver behind chamber design is mixing. Complete combustion of the propellants is the ideal situation for any rocket, but is rarely achieved; it would require a chamber that was prohibitively long to give the propellant enough time to combust completely. Chamber volume is really the key and is defined in rocketry by the characteristic length (see combustion chamber sec.). The characteristic length is equal to the volume of the chamber (injector to throat) divided by the throat area. Different propellant combinations require different residence times and, therefore, different characteristic lengths for optimum performance. These values are typically in the realm of 50-150 cm, but actual engine designs tend to use slightly lower values, ~40cm.
In regards to flow properties, the chamber holds high pressure subsonic gas, which is pretty unpicky when it comes to duct geometry; there are practically no losses due to the contraction to the throat, and the geometry doesn't need to be subtle or particularly smooth (this being the geometry of the "reverse thrust" section in your drawing). Combustion chambers are essentially fancy tanks as far as geometric design is concerned, and there's not much performance increase to be yielded from changes in that design. The nozzle design, on the other hand, is very critical. Supersonic flow is very picky, and pressure losses are easy to collect if the geometry doesn't smoothly expand the flow.
In regards to your drawing, the "backward thrust" section is entirely canceled by the forward thrust on the injector wall that is directly behind it. This leaves only the forward thrust from pressure on the nozzle and from a throat sized patch on the injector (everything you could see looking into the butt of an engine). This is further evidence that chamber design doesn't significantly impact thrust, at least in the way you're suggesting.