variation of acceleration magnitude with bandwidth

I am trying to observe the acceleration caused due to a deterministic sudden impact with an ADXL375 device.

I notice that the peak acceleration being reported varies with the bandwidth setting. The value gos as high as 150g at a BW of 400Hz, and the same impact drops down to about 60g when sampled at 100 Hz.

Is this variation a result of the decimation being performed to down sample the acceleration signal to the desired (lower) bandwidth?

How can I tell what is the "correct" peak acceleration based on these readings?

The width of the acceleration spike at impact was measured to be of the order of 10 millisec.

Thanks,

Ben