In this paper, we study multiple-input single-output joint processing (JP) systems with limited feedback where two adjacent base stations exchange both channel state information and their data. To optimize the sum-rate performance of the JP system, we propose a new feedback bit allocation method which maximizes quantization accuracy in the presence of pathloss. The quantization accuracy is formulated by the expectation of the inner product between the actual channel vector and the quantized channel vector. In order to maximize the quantization accuracy, we employ a new method which compensates the phase difference of the two channels. Through numerical evaluations, we show that our proposed feedback bit allocation strategies provide about 50% performance gain in terms of the sum rate performance compared to the conventional method with the equal bit allocation scheme.