The semiconductor laser is commonly used as a light source in fiber-optical telecommunication systems. In order to send as much information as possible in a short time, it is important that the laser has a large modulation bandwidth, i.e., the turn-on and turn-off time should be as short as possible. In analogue fiber optic systems for transmission of radio or television signals, it is also important that the light from the laser increases linearly with driving current even at high modulation frequencies. Otherwise, the transmitted signal will become distorted. The modulation bandwidth and the modulation distortion are dependent both on the laser structure and the gain characteristics of the active material. One of the most useful approaches for the time-domain description of the response of optoelectronic devices is the so-called "rate equation model," which has been widely used to describe laser performance. Commonly, laser models with simple gain expressions are used for simulation of laser dynamics. In these models the small-signal dynamic parameters like the differential gain and gain saturation parameter are extracted from modulation response measurements. However, we show that in order to correctly calculate distortion, an accurate model of the dependence of gain on carrier density, n, and photon density, s, is needed. Commonly used gain models, fitted to give exactly the same modulation response can give significantly different distortion behavior.