pfft~ and framedelta~/phasewrap~/frameaccum~

I have been analyzing and recreating the more basic patches by Jean-François Charles for doing some spectral freezing effects. However, there are a few objects that I just cant understand when reverse engineering the patch:

Before loading the polar co-ordinate pairs into the matrix, the phase/angle values get run through a framedelta~ and then a phasewrap~ afterwards. Then when they get read out the other end, the phase vales are ran through a frameaccum~. Why? Why not just store the values as is? This seems to be central to the patch though, because if you didn't need to then you could just spit the real/imaginary pairs into a matrix and wouldn't even have to convert them.

ps. excuse my ocd color coded commenting. It's uglier than Gary Busey but it helps!