You should not try to quantify RNA on agarose gels (if you are using standard dyes like ethidium bromide). RNA does not bind these dyes like DNA does. For example, ethidium bromide is generally an intercalating dye. Since DNA is double stranded, the longer the DNA molecule, the more dye will bind, and hence you get more fluorescence. There is generally a linear relationship between length of DNA (as long as it is not supercoiled) and amount of fluorescence. This is not true for RNA, which is only partially double stranded due to secondary structure. The amount of fluorescence is determined by the amount of double stranded regions, which is very sequence specific.   If you have 1000 nt of poly G, you will get almost no fluorescence, whereas 1000 nt of alternating CG will give plenty of fluorescence. So bottom line, the Qubit or nano drop are far more accurate methods to determine RNA concentration. Now, why the difference in the qPCR results? First, they are different samples and may have different contaminants. However, you column purified your samples, so it is unlikely you have such different contaminants between the two samples. I assume you are reverse transcribing the RNA prior to PCR. Reverse transcriptase can have different efficiencies on different RNA molecules. One RNA molecule may have a lot more secondary structure to plow through than the other molecule (as suggested by the different appearance on the agarose gels). Just adding equal amounts of RNA does not mean you will get equal amounts of cDNA out of the reaction. If you think differences in RT of the RNA is the problem, then do RT first, purify and quantify the cDNA, then add equal amount of the cDNA to the qPCR reactions. 

You should not try to quantify RNA on agarose gels (if you are using standard dyes like ethidium bromide). RNA does not bind these dyes like DNA does. For example, ethidium bromide is generally an intercalating dye. Since DNA is double stranded, the longer the DNA molecule, the more dye will bind, and hence you get more fluorescence. There is generally a linear relationship between length of DNA (as long as it is not supercoiled) and amount of fluorescence. This is not true for RNA, which is only partially double stranded due to secondary structure. The amount of fluorescence is determined by the amount of double stranded regions, which is very sequence specific.   If you have 1000 nt of poly G, you will get almost no fluorescence, whereas 1000 nt of alternating CG will give plenty of fluorescence. So bottom line, the Qubit or nano drop are far more accurate methods to determine RNA concentration. Now, why the difference in the qPCR results? First, they are different samples and may have different contaminants. However, you column purified your samples, so it is unlikely you have such different contaminants between the two samples. I assume you are reverse transcribing the RNA prior to PCR. Reverse transcriptase can have different efficiencies on different RNA molecules. One RNA molecule may have a lot more secondary structure to plow through than the other molecule (as suggested by the different appearance on the agarose gels). Just adding equal amounts of RNA does not mean you will get equal amounts of cDNA out of the reaction. If you think differences in RT of the RNA is the problem, then do RT first, purify and quantify the cDNA, then add equal amount of the cDNA to the qPCR reactions. 

Very interesting and informative answer Prof David.

- Can safe dyes such as gel red or cyber green act differently or give better result after agarose gel electrophoresis?

- I agree that longer DNA molecules goes with better fluorescence but, can higher fluorescence in short DNA fragment (as compared to large DNA molecule) after a restriction digest be only explained by DNA fragments of the same size?

- Is RT-PCR setting largely differ from normal PCR? I mean can secondary structures not be removed by denaturation?

Thanks for your explanations once again (sorry for my approximative english :-)    )

For Syber green and gel red stains, it depends on their mode of binding. If they intercalate into double stranded regions, then they will behave like ethidium bromide. Both ethidium bromide and Syber green can bind to RNA without intercalation, but it is much less efficient. In those cases, there is not a linear relationship between length of RNA and fluorescence. This is why, in general, agarose gels should not be used for quantification of RNA. 

With DNA molecules, there is a linear relationship between DNA length and fluorescence. After a restriction digest, the shorter molecules will have less fluorescence than longer molecules. There are some anomalies that have to do with homopolymer tracts, but I don't remember the details. 

Secondary structure is removed by the denaturation step in PCR. But since secondary structure formation is a reaction with zero-order kinetics, the secondary structure in the RNA can re-form almost immediately. With DNA in the PCR reaction, denaturation is first order kinetics, so renaturation takes much longer.

in zero order kinetics, the rate of production of the product (the secondary structure) depends only on the concentration of reactants (the denatured RNA).

Thanks to much for this helpful details.

Try a no exogenous primer control in your cDNA synthesis reaction and no primer control in the PCR  Self-priming is a known phenomenon and can lead to differential cDNA synthesis efficiency.