Wyner considers fixed user location and homogeneous interference over the cell, i.e., interference doesn't vary within the cell. For a user in a network, it receives signal from all the BSs in the netwotk. The received signal Rs= GcRt+ noise. Gc= channel gain. Rt=trans signal

The channel gain for the serving BS would be comparitively more than the interferers. It can not be more than 1 as it wont boost the signal. The serving BS gives the best possible signal. Since the channel gain varies with user location, fading and other channel parameters,  wyner considers fixed user location and doesn't consider practical constraints,  therefore, channel gain is assumed to be unity from serving and less than unity from all other BSs. 

Thanks #MP