RT-PCR: inconsistent reference-gene replicates

[Irideus]

Hi everyone.

I´m trying to normalize the level of amplification of a fungal reference gene (alpha-tubulin) among five different cDNA samples taken from inoculated susceptible plants. This is the first step to quantitate the expression level of several genes-of-interest.
I have roughly calculated the efficiency of the PCR reaction by doing densitometrics on the bands obtained at different cycle numbers (e.g. 20, 23, 26, 29, 32), and selecting cycle numbers from the linear amplification phase. I did this in order to be able to calculate the No (number of input molecules) of each reaction, out of the intensity of the product band.
For example, if the No of cDNA sample X is 100, and No of cDNA sample Y is 200, I dilute cDNA sample Y 1/2, in order to get the same amplification level for both CDNA samples after a given number of cycles.

The problem is that after making the dilutions according to the densitometric data, the intensities of the bands change their proportions again. Even if I achieve the same intensity in a given pair of cDNA samples, this is not reproducible. I also ran triplicates (out of the very same PCR mix for every cDNA sample) and the intensity of the bands are still inconsistent.

I have a good deal of experience doing PCR. Still I definitely have no clue of what is wrong with the replicates. I COULDN´T be more careful in my pippeting... I thoroughly mix my tubes... I changed the Buffer, dNTPs, MgCl2 stocks... I keep them on ice all the time... etc., etc.,... and still the same...

Maybe fungal RNA is so scarce in my samples that stochastic effects are preventing replicate uniformity?
Maybe the amplificaton efficiency (around 10%) is too low (thus making the stochastic problem even worse)?

I´ll deeply appreciate any suggestions (apart from dropping career, of course).

Thank you for reading.