Thank you very much for your reply. That is just from the knowledge of the real scenario and application. Although, it might not be clear in this illustrative example, imagine you are going to cluster a set of people into different groups. Then you should have some idea as to how many people must be in a group. If you pick only 2 nearest neighbours (other than the point itself) you are likely to obtain clusters of around 2 points or lonely points. However, if you tolerate distances up to 10 nearest neighbours, you could reasonably assume that you accommodate enough noise to contain a larger number of people in a group.

I hope this helps.