Confusion about expected number of CNV calls

[tsin-lau]

Thank you for developing this excellent tool for CNV calling and it works well with my WES data. However, I'm still confused about the number of CNVs of each exome.
As mentioned in ExomeDepth-vignette, you usually obtained 150-280 CNV calls per exome sample. But according to the reference manual, the argument, transition probability, in the CallCNVs function is described as follows —— Transition probability of the hidden Markov Chain from the normal copy number state to either a deletion or a duplication. The default (0.0001) expect approximately 20 CNVs genome-wide.
I wonder where the discrepancy comes from and how the transition probability argument can be modified.
THANKS!