You have to know the nature of your design spectrum. Is it a geometric mean of two horizontal components? Or is it the maximum rotated component? Most spectra for design have historically been geometric mean. Bridges in the US still use geometric mean. Buildings have moved to maximum component.
This issue is one of those discussed in the following paper and a method is proposed for determining component target spectra for 3D modeling.
"Issues in the Prediction of Inelastic Behavior in Bridges during Earthquakes"
You need to come up with a relationship between max component and min component. Let's say you have a geometric mean design target spectrum and decide that you will take the max component equal to 1.2 times the geometric mean. Then the min component would be:
SQRT(min x max) = geomean SQRT(min x 1.200geomean) = geomean --> min = geomean/1.200 = 0.833geomean
So you would match one component to 1.200 times the target and the other component to 0.833 times the target.
This is where the spectrum multiplier in SeismoMatch will come in handy. You don't have to redefine the target spectrum, just change the multiplier.
This is one of the reasons that scaling is typically preferred over spectral matching. In scaling, there is no need to assume a relationship between the components.