The analog mean-delay (AMD) method is a new powerful alternative method in determining the lifetime of a fluorescence molecule for high-speed confocal fluorescence lifetime imaging microscopy. Even though the photon economy and the lifetime precision of the AMD method are proven to be as good as those of the state-of-the-art time-correlated single photon counting method, there have been some speculations and concerns about the accuracy of this method with respect to the absolute lifetime value of a fluorescence probe. In the AMD method, the temporal waveform of an emitted fluorescence signal is directly recorded with a slow digitizer whose bandwidth is much lower than the temporal resolution of the lifetime to be measured. We have found that the drifts and the fluctuations of the absolute zero position in a measured temporal waveform are the major problems in the AMD method. We have proposed electrical and optical referencing techniques that may suppress these errors. It is shown that there may exist more than 2 ns drift in a measured temporal waveform during the period of the first 12 min after electronic components are turned on. The standard deviation of a measured lifetime after this warm-up period can be as large as 51 ps without any referencing technique. We have shown that this error can be reduced to 9 ps with our electronic referencing technique. It is demonstrated that this can be further reduced to 4 ps by the optical referencing technique we have introduced.