I am not sure your way with constant Dispersion actually works well.
You are using 16ps/nm and 0.08slope, and for DCF -80 and -0.45 slope. Suppose you have one signal 1nm away from the reference: in 5km propagation its dispersion is 80.4, while its compensation is -80.45. I am afraid that at high bitrate the 0.05 difference may affect the performances…