The Bullet Cluster result remains quite striking, due to the relative simplicity of its structure. There are many other systems which are more difficult to understand.I'm not saying it does. I am saying that a single observation (in this case the Bullet Cluster) is not nearly as simple and as unambiguous as the popular press (and some astronomers) make it out to be.
But I still think the CMB is vastly stronger evidence for dark matter. The Bullet Cluster's main points in its favor are that it's easy to explain and has some pretty pictures associated. Even though its structure is relatively simple, any galaxy cluster will remain pretty complicated.
But with the CMB, the physics involved are tremendously better-understood than the physics responsible for the formation and behavior of galaxies and galaxy clusters. The fundamental quality that makes the CMB so easy to deal with is that the CMB was emitted long before most of the matter in the universe collapsed into dense configurations (e.g. galaxy clusters). The smoothness of the universe when the CMB was emitted makes it so that linearized gravity is an excellent approximation. With the ability to use linearized gravity, it's possible to calculate very precisely the behavior of the early-universe fluid given the physical properties of that fluid.
This contrasts with the formation of galaxy clusters and galaxies, where many of the calculations are simply impossible to do in an exact manner. We have to instead use approximations such as N-body simulations to attempt to determine their behavior over time. To get a rough idea of how much easier it is to calculate things using the CMB, it's possible with a modern desktop computer to calculate the power spectrum given a model of the universe within seconds. N-body simulations generally still require large clusters to run the simulations over an extended period of time. There are ways to make approximate calculations for large-scale structure which take less time, but they always sacrifice significant accuracy.