Adaptive network coded cooperation (ANCC) provides a general coded cooperation framework to map the communication connections of a wireless network into the coding constraint. However, random selection policy is assumed at each node to select its retrieval subset, which cannot avoid illness in code graph at the sink. In this paper, the concept of indicator matrix is proposed to regulate the retrieval subset selection to derive the enhanced ANCC (EANCC) scheme. It is unveiled that the indicator matrix based retrieval subset selection policy can effectively tailor the girth and degree characteristics of the resultant code graph. Since the most harmful short cycles in the network coded cooperation can be completely eliminated, the proposed EANCC approach promises better error performance than the existing ANCC scheme. Numerical results show that it performs 13 dB better than the ANCC scheme at the bit error rate performance of $10^{-5}$ in seven-user network over the Rayleigh block fading channel.