Ironically, you might reduce the efficiency of the compression, because you have squeezed out the "redundancy" upon which text compression algorithms depend. Simply stated, these algorithms recognize "a sequence that we have recently seen before," and replace them with a coded reference to that sequence. Your pre-processing might remove that redundancy, reducing or even eliminating the chance to substitute codes. However, there's really only one way to find out: an empirical test. Generate a few thousand bits of plausibly representative data (unless you have the real stuff at hand), and try it both ways.