Cirrus has a list of deployment events on their site. The failures mostly appear to be cases where the parachute was deployed too low.

https://www.cirruspilots.org/Safety/CAPS-Event-History?id=3

It's not 100%. In one of the flying mags you can read a story about how a VFR (not instrument rated) pilot got into the soup, lost attitude awareness, freaked out and pulled the chute lever. Nothing happens. While the pilot was yanking on the chute lever, taking hands off the stick let the aircraft's static stability take over and the plane flew out of the cloud by itself. The pilot then took over and landed in the usual manner.

That is not a correct interpretation of that statistic.

The parachute has never "failed" in any engineering sense. The stat is taking into account all deployments, including those well outside of the deployment envelope, such as not enough altitude or too much velocity. No one can expect any parachute to deploy if you're too close to the ground.

Within the envelope of deployment, the statistic says it has a 100% success rate.

the chute has deployment parameters (altitude, airspeed of plane, etc) that aren't always met by the pilot in an emergency.

If the pilot flies into terrible weather or gets into an uncontrolled spin, it's not reasonable to expect the chute to deploy correctly and save the situation.