How do I get peak deceleration from an acceleration graph?

Firstly, your intuitions are correct. You find the peak deceleration (which I will hereafter refer to as acceleration, because "deceleration" is just a term for acceleration in an arbitrary direction which you consider to be "backwards", or of negative displacement) from the first peak; but not so fast - it is the value of the first peak, minus 1 g, assuming you are interested in the outside observer's frame of reference.

This is because the accelerometer seems to be measuring the gross acceleration upwards on the ball, rather than the resultant (net) acceleration.

For example, when the ball is at rest, it seems to measure an acceleration of about 1 g - you can think about this in the sense that the upward force on the ball due to forces that are not gravity is equal to the weight of the ball; as it is continually at rest. (The electrostatic, repulsive, normal reaction force must be equal to the weight of the ball to keep it in one place).

The acceleration is only zero g when the ball is falling - as there is no upwards force on it. You can consider the graph, then, to be shifted upwards by 1 g. The deceleration you want would be the peak of the bounce, minus 1 g.

However, keep in mind that taking the raw spike height may be correct as well, depending on what frame of reference you want to take. It is more likely, for elementary physics, though, that you would want the value corrected by a term of 1 g.