The Power Accuracy of +/-3% refers to the absolute uncertainty of the measured value. For example, for a 2 Watt reading, the actual "true" value would be between 1.94 W to 2.06 W (with reference to NIST, to which all our calibration is traceable). This assumes the reading is from about 5% of full scale up to full scale. It should be noted that our accuracy specification is in general based on 2 sigma standard deviation.
Repeatability of the measurement (assuming the laser itself is perfectly stable) is limited in the best case by the power noise level of the sensor, and is typically better than +/- 1% depending on the thermal stability of the environment. Stability at higher powers from the middle to the top of the range of the sensor head is usually better than the low end. This is due to small temperature variations having less of an effect as they are proportionally a lower percentage of the total power. For more information, refer to our Web tutorial at: https://www.ophiropt.com/laser-measurement-instruments/laser-power-energy-meters/tutorial/calibration-procedure