I worked with a lesser model of Analog Devices gyroscope. They aren't bad, but BGAs are evil to hand-solder.
The 25 degrees C is a point of reference for the temperature output. The datasheet says that at 298K (which is 25C), the voltage output on the temperature sensor is 2.50 volts. Useful if you're looking for ultra-precise measurement.
As for calibration, the gyro module itself has no calibration capabilities other than to offset the 2.5v reference voltage (I wouldn't recommend that). If you're getting discrepancies between the axes, then I'd say make sure the voltage on the reference line has no interference - put a large capacitor in parallel with ground, or put the device on an independent circuit. If all else fails, try to calculate an optimal scale factor for your circuit - iMEMs devices, as with all sensors, have slight differences inherent in manufacturing.
Good luck
Sparks
EDIT: Keep in mind that the scale factor for one gyroscope may not work for the others on that board. Also, looking at the
datasheet shows you that the temperature has an effect on the output - about 8.4 mV/K - which could account for your discrepancy.