I do not see why saliva samples should be different than other bodily fluid samples I have used in research. The first two things to look at are 1) are the reading you obtain at 562 nm within the linear range of your standard curve for all dilutions? some may be too high and some may be in the lower end and thus the results obtained not reliable; 2) are you using the same amount of the reagent in all the dilutions? As the instructions state, the BCA reagent should be at least 90% of the total assay volume and the other 10% should be "sample + water." If the amount of BCA reagent is different, then the 'blank' used to zero the spectrophotometer will not apply to all samples and cause different values for different dilutions. Your 'blank' should also be 90% BCA reagent + 10% water. And remember, all assay tubes must add up to the same total volume.

I hope this helps.

Dear Eduardo,

why do you use triplicates instead of more dilutions? Remember that you are using curve fitting for your dose-response curve (ideally you'd get one that has a nice linear range as Gustavo pointed out). The linear regression will also ensure proper weighting of the data points. Do this for the standard curve and your samples! Do not calculate the concentration of your individual diluted samples and multiply with the dilution factor and then calculate an "average". This way you'll amplify small errors at high dilutions (low readings) and this will affect your overall results.

If you have a linear range you can use the following formula for calculating the concentration of your sample:

c(sample) = slope(sample) / slope(standard) x c(standard) x starting_dilution(sample).

Feel free to post the data set so it can be evaluated by me / us.