[Al Skierkiewicz]

Not to get confusing to new kid on the block, but motor curves are given for a fixed 12 volt hard regulated source voltage. Then the manufacturer varies one of the other variables to obtain the curve. (See the CIM motor curve in the FRC Suggestions Document) Note that speed, power, current and torque vary in relation to each other. Add more load or demand more torque and the current rises and speed falls. Reduce the load and the current falls but the speed rises. Design the load so that your application will keep the motor in one area on the curve and you can be sure that the current will also remain in that area of the curve. Ask the motor to deliver too much torque and the current will go sky high.
In our controllers, the PWM nature of the output simulates a change in current and therefore speed and torque are affected. Even if the load is relatively constant, the change in current moves the operation of the motor with repsect to the curves. However, efficiency and peak power cannot be determined mathematically from no load speed and stall torque. Nor can no load speed be calculated from the tables. As Russ has pointed out, Back EMF affects the motor response. If a motor has all load removed, then the free speed is determined by the back EMF. At some high RPM (which varies with the motor design), the back EMF and supply voltage cancel each other and the motor speed can go no higher (for a fixed power supply voltage).
The controller (Jaguar or Victor) is always switching the supply voltage less any series resistance between the battery and the motor. You will note that in all of the curves supplied, the specifications are always given at 12 volts DC. Take a look at the file below for the 3.5" CIM motor that was a KOP motor in 2006. You will see that max power occurs at the point where the current and efficiency curves cross. But note where the max efficiency actually occurs, at a much higher RPM and lower current.
So as I have stated in other forums, although the output change per PWM value is more linear for the Jaguar, talking about linearity when connected to a real world motor, means very little. Your load, series wiring, software and motor will all be different from all other users. The battery voltage changes over a match and with how many motors are turned on and their respective loads. Every motor we use on the robot has a varying load and that varying load will move you around on the motor curve. Turn on 4 CIM motors and drive into the side of the field, and the power supply will fall. If you are still giving the same PWM value to a motor controller, it cannot drive the motor speed to the same point as when the four CIM are not being driven. Linearity only holds true when all other variables are held constant.