Enhance Speech Creates Weird Audio Spikes

Something I've noticed semi-consistently is when using enhance speech in the essential sound panel (which 95% of the time is amazing) is that it will create words or sounds out of either a background noise, such as an object being tapped, or a breath by the talent.  I know the machine learning is still coming along, but would like to see this addressed in the future so I don't need to cut the audio separately to remove the enhance for that instance and try to replicate the enhancement on the rest of the clip.