PD measures localized loss relative to diffuse loss (i.e., generalized loss or generalized depression). It's essentially local loss minus diffuse, or general loss. If a defect is localized and there is not much diffuse loss (e.g., disease is early to moderate, with little diffuse loss), you'll get a high PD value. If glaucoma is so far advanced that the diffuse loss is as deep as any localized loss, you'll get a PD close to 0. This is why PD sort of "corrects for cataracts". When you subtract the loss caused by cataract across the measurable visual field, you then can detect localized change. I think this makes sense...
Several years ago I found a sistematic error (with Octopus perimetry) in patients with advanced glaucoma (published in Perimetry Update 1998) because there is a sistematic progression of false negative rates. It needs a correction (you can see the manuscript in my profile). So, MD and PD must be analyzed in this perspective.