I am working with a similar kind of network. I don't know if it is correct, but I think the most proper way is to consider the SNR from the first transmitter, not considering what the relay do to the signal. This because you want to analyze the performance of the end-to-end communication. SO you send the signal with some energy, and you want to know the performance of your signal once you have sent it with that energy.
That's the way I am doing, and comparing it to the direct transmission, I am getting a very similar performance: a difference of less than 1dB for the same BER.
I am not amplifying the signal in the relay, so I am just doing the decode and forward (DF). I think it is reasonable to get this "worst" performance in the DF relay compared to the direct transmission because you have two channels, therefore you have two components of noise. You perform the DF in the relay, but it can decode wrong.
Let me know if this answer was useful, and if you have any more questions. I am not an expert in this area, but I am working on it.