Al-Kisai learned Hamza's reading before he had one of his own, and the relationship between the two is the clearest illustration in the canon of what a qira'a actually is. He did not invent readings. He did not depart from the consonantal skeleton the Uthmanic codices fix. What he did was choose, at each point where the received material offered more than one defensible way of reciting, and the set of his choices is his reading.
Where he agrees with Hamza he agrees because he learned it from Hamza; where he differs, the alternative he takes is one that other named teachers in the Kufan line were already transmitting. This is why the discipline speaks of a reader ikhtiyar, a selection, rather than of a composition. It also explains a feature of the canon that can look untidy from outside. Readers overlap heavily.
The Kufan three, Asim, Hamza and al-Kisai, share a great deal of material and are distinguished by a body of specific differences rather than by three separate texts. The same is true of the two Medinan readers and the two Basran ones. A canon of ten readings is not ten versions of the Quran. It is one text with a bounded and fully enumerated range of accepted recitations, and the boundaries of that range are what the readers and their chains define.