Thanks for that link - I'm still slightly confused, as the schematich there uses an additional "tri-state switch", which I didn't find necessary. But poking arround that website gave me this:
http://www.barello.net/Papers/Motion_Control/index.htm
There is an alternative way to drive the motor: The Enable bit is left ON all the time. One half of the H-Bridge is set forward or backwards and the PWM signal drives the other half of the bridge. The effect of this technique is that the motor is alternatively driven and shorted with the ratio set by the PWM signal
which I guess is what I'm doing. It even hints at using "locked-antiphase" to use only one pin per motor, but that I haven't understood quite yet.