[jgannon]

I'm going to have to agree with everything above. I found it easier to reverse the leads on one set of CIMs, rather than having to fix it in code. Also, the CIMs do run almost identically in forward or reverse, so I would say that there must be something whack with your right tranny.

<edit>
In an attempt to actually add something new and insightful to the discussion, rather than just repeating what everyone else said, here's how I fixed the drift on our drill motors last year (which do indeed run very differently in forward and reverse). We hooked up an encoder to each wheel (we just used Banner sensors and white dots on the wheels, since it was far too late to integrate a real encoder). Then I wrote a segment of code in Default_Routine() that ran the drive motors for about ten seconds (350 program cycles, I think), counting encoder clicks the whole time. I'd put the robot up on blocks, and then execute the code with a specific combination of joystick button presses. At the end, it would determine the ratio of encoder clicks (a/(a+b)), and then apply that difference to the faster motor. Basically, instead of hard-coding the ratio as you've done, you could do this to adjust the ratio on the fly, and store the result to EEPROM. Then you can recalibrate it whenever your drivers start to complain again. See if there are any obvious problems with your tranny first, and then try this.
</edit>