Internal control for absolute quantification - (Aug/06/2015 )

I use real-time PCR to measure plasmid DNA concentrations (fg/ul) in my samples. As a reference standards I use 10-fold serial dilutions of  plasmid pBR322 (from 5 to 500 000 fg/ul). My standard curve has good slope and efficiency, but I noticed that sometimes the measured concentration of the same sample differs in different experiments with different standard curves. That is why I decided to prepare an internal standard with concentration of 10 000 fg/ul. But I have problem with preparing these standard. I observed that once the obtained concentration is near 10 000, bo sometimes it is 5 000 or even 25 000. I prepared these solution in 50 ml with different startegies: serial dilutions or by adding concentrated pBR 322 directly to water. What might me the reason of such differences? Could you please recommend me how to prepare such solution? Thak you very much for your help.

You use a nanodrop to measure your concentrations?

Only for higher concentrations: 5x10^8 and 5x10^7 fg/ul - NanoDrop isn't sensitive enough to measure lower concentrations.

There could be a number of things here.
You say you're preparing standards of 5 to 500,000fg/ul.   Are ALL of those for your standard curve?   Your standard curve should bracket the concentration of your unknown samples but not by any more than 10x the concentration of the highest value (and that's an absolutely highest limit.   You should more aim for 2x the highest unknown).
Your standards all need to be manufactured and tested via the same method.   If this is not practical then you really need to understand the uncertainty of measurement for each method.

You say that your standard curve has a "good slope and efficiency" but you don't detail what that means.   Are you using the coefficient of determination to measure your standard curve?   How close to 1.0 must the curve be?   Also, be VERY careful when analysing standard curves and do not rely only on the numbers - look at the curve for goodness of fit.   I have always used Anscombe's quartet to illustrate this (Google it).   It doesn't matter that AQ uses correlation coefficient and not the coefficient of determination, it highlights the same thing.

So, if you have a R² value of 0.95 and you're straddling an enormous value of DNA concentration then very small deviations from 1.0 can conspire to have a very large effect on your calculation.

I'm not a massive fan of Nanodrop but I wonder what you're using for the lower concentrations.   You may be exerting enough leverage on your curve that calculated values in the middle of the curve are pulled way out of true.

All this aside, I'd question how you are conducting serial dilutions.   Do these with as large volumes as possible with as accurate and precise instruments as possible.   There's no point messing about with 1ul volumes with a p-20 pipette.    And use the same pipettes throughout for all parts.   That way you reduce stochastic errors and keep systematic errors uniform throughout your process.
Also, if you are not 100% confident in pipetting techniques, ask someone who is.   Pipetting well consistently is a skill that is worth its weight in any precious metal you wish to trade.