How are you excercising the channel in question ?

I remember that when I got my first high-spec servo (digital, coreless), I hooked it up to the aileron channel of my RX to test it, thumbed the stick all the way over and watched the servo move, then when I let go of the stick and watched the servo go back to centre, I though something was wrong since the servo did exactly what you describe - it went past centre, back again, oscillating very slightly before stopping.

Took me a moment to realise that the problem was that I was letting go of the stick and the springs on the stick were actually making it do this damped-oscillation ; the only reason I had never seen this before was that the cheap and slow analog servos I had used before had so much inertia that the servo itself damped out the momentarily varying position that was being sent due to the stick position settling down.

Do your tests with you keeping control of the stick at all times so that the springs aren't bouncing it around ... or connect the servo to your retract channel or similar, where there is no such problem of spring-induced variation of the servo position info.