Real-time standard curve amplification problem?

[bdietzpu]

I have made a dilution series for my standard curve 1:2, 1:4, 1:16, 1:64, and 1:256. The problem is that my 1:256 dilution almost always amplifies very late and the 1:64 dilution usually does as well. I have tried several different sets of cDNA and had the same problem. When analyzing the melt curve, all of my samples melt in a very tight range, except for these failed standards, which come up way before the others. I have made straight dilutions (not serial) from the same cDNA and they worked perfectly, but introduces too much pipetting error to normalize my results to. It is my understanding that in the past, this machine had no problem analyzing these low dilutions and the curve could even be extended out one more point. I have OD'ed my standards on the nanodrop and they each contain a 4-fold difference(64ng/ul, 16ng/ul, 4ng/ul, etc) so I don't think that I have any DNA degradation. The melting curve makes me believe that the problem lies with the primers and I am just seeing primer dimer, but then why is the serial dilution failing when a straight dilution does not? If anyone has ever had a similar problem and found a solution, or has any idea what could be causing this frustrating problem, I would really appreciate the help.