When you choose "Ratio to heavy" as the normalization method, Skyline expects that the heavy standard will be present at approximately the same concentration in all the samples and the light concentration will be different for different concentration levels.
For this reason, "Ratio to heavy" is not the correct thing to choose if your calibration samples have different heavy concentrations.
Your calibration curve has a negative slope because the heavy concentrations are changing in the completely wrong way in your external standards for what would be expected if the heavy peptide were being used as an internal standard.
Skyline does not really support "reverse calibration curves" in the sense that you are thinking of them.
People do use Skyline to create response curves using spiked in heavy standards, but the only purpose of those response curves is to figure out what peptide concentration is within the linear range of the mass spectrometer.
After you have verified that the peptide that you are looking at is in the linear range, then you would probably spike your heavy standard into your unknown sample at a known concentration, and then just look at the "Ratio to heavy" value to figure out what the analyte concentration is.
You can also use the Document Grid to fill in the values for a column called "Internal standard concentration" and then the "Calculated Concentration" value will be the ratio to heavy multiplied by the Ratio to heavy value.
Here is a support request where someone asks about reverse calibration curves and the answer was that the only thing you should use them for is verifying the linear range:
https://skyline.ms/announcements/home/support/thread.view?rowId=36322
If you would like to learn more about the Document Grid this is a good tutorial:
https://skyline.ms/wiki/home/software/Skyline/page.view?name=tutorial_custom_reports
-- Nick