How are error rates in voting method of ensemble learning calculated




While learning about ensemble learning(voting) I cam across the following:

I could not understand what is being meant by the highlighted part in yellow.If I am not wrong,since here there can be only two outcomes-vote is right/wrong the distribution is binomial but why is the p taken to be less than 0.5?Is that the maximum permissible error rate or something??
Also the chart and it’s explanation is not clear to me.
Can somebody please help me with this