the derivation of lambdaModeDecisionLdSad #636

When I derived the constant array lambdaModeDecisionLdSad, according the comment of JCTVC-F802, the data couldn't be fitted perfectly. For example, lambdaModeDecisionLdSad[0] = 12456, from JCTVC-F802, lambda = int((sqrt(a * W * power(2, (qp-12)/3.0)) * 65536), whether W is 0.57 , 0.68 or 0.85, a is 1 or less , the result didn't agree with the array. So, is there anything incorrect in my derivation process. I would appreciate it if you could offer me some guidance .