Provided the on-current is within the maximum rating of the peltier device, that's true but the PN junctions do exhibit the forward current characteristic of a diode with a small ohmic series resistance. There's usually a few hundred of these junctions in series (but in parallel for the thermal path) so a 6A 12v device may have an effective total resistance of 2 ohms (the forward volt drops at each PN junction are only a few millivolts each).
As you increase the current, the I squared losses eventually start adding more heat than the device can pump from the cold to the hot side. If you set your on-current setting of your PWM controller to this critical limiting value, no matter the duty cycle, each current pulse will add as much heat as the junctions can transfer. This is an extreme example of why you don't want to drive a peltier device directly from a PWM current source.
And others that say 50Hz is fast enough (but only from the point of view of thermal cycling stresses, ignoring the other implication of using a high duty cycle PWM supply). Cycling the device 2 or 3 times a minute accelerates thermally induced fatigue stress. The best way to control the cooling is to regulate a totally smoothed DC current supply (for best efficiency, by using a switching regulator).