In my opinion, it doesn't matter to put independent values on X axis and response value on Y axis always. I think It's depend on the way that you want to present. For ELISA curve you either set the Standard concentration value on X axis and OD value on Y xis or vice versa. The curve can display the equation of the curve to you. Just make sure the term or the variant of number that you put in the equation for response value calculation.
You know in scientific field, precision is very important. Precision depends on repeat ability and reproducible items. In this approach universe expression and calculation are very important. If you remember from mathematics and Cartesian coordinate system (X and Y axis), Mathematicians at first and then physician and chemist, accept variable (X) on horizental axis and function (Y) on vertical axis. This agreement developed in biological field and biochemists and immunologists accept this rule and agreement for drowning various types of standard curves such as ELISA method (variable on X and function or signal on Y axis). Finally after mathematicians, physicians and chemists, this is universe agreement for X and Y axis for showing variable and function respectively.
So revers manner (Variable on Y and signal on X) has no logical or rational error, but it is disagreement.
If you want to generate a calibration curve by automated means, it's easier to plot concentration as a function of OD, as software (e.g. LibeOfficeCalc or Excel) then will give you the equation to calculate your unknowns directly. If you want to show dose-response, then plot OD as a function of analyte concentration.