We actually have an interaction sprint that recently wrapped up and will be published as a blog post in the next few weeks that tackles this issue. In the meantime, I would definitely check out the projective interaction method we used in the UI Input Module: https://gallery.leapmotion.com/ui-input-module/
This works by generating a raycast from the user's shoulder, through a pinch, towards any object. It allows for easier and more fluid targeting.