I think the problem there is that it is just poorly explained.
Take for example the attached program that compares "normal" and "kashida" 'Raħīm' taken from [1], you'll get:
c1 = c2 c2 = c1
meaning that "normal" and "kashida" are the same string. So, probably that phrase should read something like: "If an application calls the function to create a sort key for a string containing an Arabic kashida character, the function will ignore that character and no sort key value will be generated for it."
