I am having problems determining weight percentages of solvent blends. I am new at my current job a was not here when the GC was set up, it is an SRI 8610C. I do not think the problem is with the GC but with the procedures that are set in place for it's operation. The current method for determining the weight percentages of components in solvent blends is to take the area% directly from the chromatograph. The method worked fine until we ran into a four component blend and then we realized that there is a problem. We do not have a method in place for extrapolating the area of the peak to determine the percentage of the compound. I have spoken with three different laboratories and each has given my three very different opinions. The first was to make a lab batch of the blend, run the sample take the area% and calibrate the GC to the weights that I used. The second was to make a standard, dissolve it in a heavy solvent and run the GC to determine the area. Then dissolve my sample in the solvent and use a mathematical calculation involving the density of the compounds to determine their area. The third was to make a standard that gives me the areas I am looking for and take the area of the compound in my sample divide it by the area of the compuond in the standard then multiply that by the weight that I use in the standard. Any help that I can get in the matter would be more than greatly appreciated.

Consider No. 4:
Create a true comparative standard with at least three different mixed component levels, plot a calibration curve for each component. Calculate relative response factor for each of the four component you wish to calculate weight%. Divide your %areas by RRF to find weight%.

Check your calculations by running different spiked blend samples using your midpoint or 100% target value. Your sample components should analyze reproducibly in both situations (-100% for the spiked blend).

Assuming you are using a FID detector, which are pretty fairly linear through a very large range, I would look into a calibration standard approx at the levels you are going to see. If you can run your samples neat without dilution, why add the extra effort and error. The Conc of the standard / area of std x area of unknown = conc of unknown.
If linearity is a problem, you'll have to added as many calibration standards as needed for the accuracy of your result.