Need suggestions on the class of blind adaptive filters that can be used in my situation

I have 2 satellite transponders closely situated, transmitting the same signal. At the receiver I receive this composite signal. I have to equalize it in such a way that the demodulator sees only one of the two signals from the composite signal. So, I modelled this situation as a channel with input $x(n)$ given to a channel whose output response is given by,

$$y(n)=x(n)+s*x(n-D)$$ where $$\text{0}\leq\text{s}\leq\text{1}\\ D=delay$$

Since I don't have a reference signal at the receiver, I am performing blind-equalization using a traversal structure. I know that ideally I need an IIR structure for perfect equalization. Since, the modulation schemes used are QPSK and BPSK I am using CMA (constant modulus algorithm) based zero-memory predictor for generating the reference signal to generate the error signal from the filter output. I am using LMS algorithm for updating the filter taps with mu=0.0001. The reason for choosing CMA is that it can also work on signals with frequency offsets. We have frequency offsets due to Doppler shifts in this case. I have also gone through non-linear equalization techniques like MMSE+Viterbi algorithm.

Can you comment on the strategy I followed and suggest any other ways to investigate this situation?

Thanks, JK