Your logic is a bit flawed...
How do you come to the conclusion that :
That means if the Current Source Register is once set, it can only be reset by making a reset by pulling RSTN to low? And it can't be reset by the 'init bit'?
especially as I've posted numerous examples of changing the CS reg value while the chips are running... 
I would strip down the code to it's simplest form and get 1 picture showing on your matrix to test the hardware etc. before going any further.