Preconditioning method developed by the authors was coupled with Building-Cube Method (BCM). Wall boundary treatment is a critical issue for Building-Cube Method as a Cartesian mesh method. In the paper, the classical case, flow past a cycle cylinder at low Reynolds number is used, with three different boundary treatments tested: direct data transfer, staircase treatment and immersed boundary method (IBM). Also, two sets of numerical simulations with different number of grid points are conducted to assess the sensitivity of the three schemes to grid density. The numerical results of separation vortex length, pressure coefficient distribution and wall skin friction coefficient distribution are derived and compared with those existing experimental and numerical data sets. Comparison shows that immersed boundary method is a best choice of the three schemes, in both boundary orientation and pressure/velocity estimation. Moreover, immersed boundary method is also proved to be less sensitive to grid density than the other two, since the accuracy loss of immersed boundary method case with less grid points is much less.