The problem you are dealing with is you are not normalizing / setting up the simulation correctly. Note that your Gaussian has a half-width of about 1.3 but the waveguide section of your structure is only about 0.5? Which means any field component that does not couple into the waveguide is reflected at the PhC interface. You are not getting a power of 1 because not all the power is getting coupled into that waveguide.
Scott