I don't think you're supposed to just average the two together somehow by themselves. The two signals are designed to be used in different zones and in practice have very different signal strengths at any point on approach.
From EVALUATION OF INSTRUMENT LANDING SYSTEM DDM CALIBRATION ACCURACIES [sic] by Dennis M. McCollum [PDF]:
The course pattern contains sidelobes that, without the clearance pattern, could cause an aircraft to select a wrong (sidelobe) course. Within the region of the true course (i.e., + 10 degrees about the designed procedural course), the signal strength due to the course array is greater than that from the clearance array. In the regions where the sidelobes exist, the clearance signal predominates. Airborne ILS receivers are designed to respond to the greater of the two signals.

This has to do with the background in processing these signals through analog means. Both the clearance and the course signals are within the passband of the receiver. Since this is an AM receiver and DDM is more or less a ratio of 90 Hz to 150 Hz, if the clearance or course is much stronger than the other, the stronger one prevails.
So in practice measuring the course signal DDM inside the clearance region doesn't reflect what's done in practice, and I'd guess it probably isn't even a guaranteed behavior of the transmitter. Do anything else instead in this region, like perhaps measuring just the field strength of the course signal.
But what about when slightly off the centerline? Then you'd get a mix of course and clearance signals. You could assume the ideal condition where a capture effect completely eliminates the weaker frequency. Another approximation presented in the admittedly confusing paper "Current Issues in Demanding ILS Ground and Flight Measurement Environments" by Gerhard Greving and L. Nelson Sponheimer is
\[ DDM_{total} = \frac {DDM_{clearance}* CR^x + DDM_{course} } {1+CR^x} \]
Although I can't quite make out what precisely CR and X is here.
Depending on the application, it may be more realistic to model the amount of distortion expected or permissible due to clearance band interference, especially due to multipath reflections of that signal. However, given your comments that probably won't help your specific application.