With pleasure, I have drawn the schematic for you this morning:
All channels (X, Y, I, R, G, B) share one offset buffer, and all channels use exactly the same differential amp circuit. The X, Y channels are bipolar differential, and the R, G, B, I channels are unipolar differential, but this is accomplished in software.
I suggest that you use a +/- 5V power supply with rail-to-rail op-amps. It means the output voltages are limited to safe levels without needing any additional circuitry. I used the IA0509S DC-DC module (from the original correction amp kit) to provide the +/- 5V power supply from USB, and TLC074 op-amps.
The offset adjustment potentiometer can be any value between e.g. 10k and 100k. The 47k resistors can also be any value between e.g. 10k and 100k, as long as they are all exactly the same. The 100R resistors are not critical either - their purpose is only to limit transmission line effects and short-circuit current.
I'm aware of many ways to improve this design, but I choose not to implement them because additional complexity is unnecessary. The schematic above works well.
As you can see from my screen captures in other threads, my work is in Linux. I tested the DAC using C-Media cards based on CM106L and CM6206. It works well using the ALSA-provided USB audio driver as well as my own kernel-mode driver on Linux. I also used a Windows computer to run some of the free Zoof-ware through the drlava EasyLase wrapper without any problems. What "setting may be out of whack" specifically?