In Katz & Lindell's textbook (2nd edition)) is said, that only non deterministic encryption can lead to security for multiple encryptions. Now I looked at the experiment for multiple indistinguishable security and there is said, that the challenger gets two sets of messages from the adversary. Say we have the PRG cryptosystem, that XORs messages with the output of the PRG. Why couldn't the challenger simply create a pseudorandom string equal the combined size of all messages in the set using the PRG and therefore create one big key, that is used to encrypt all messages? This should be deterministic and still fullfill the requirements of the experiment?