On the effective-medium approximation for bond-percolation conductivity