The hash solution sounds good to me. The flag solution requires code to be altered in places outside the two areas of concern - where the status is detected, and where the constraints are checked (you have to create and initialise the flag variables). With the hash you may be able to detect a little more, like duplicated data in the record (the key exists already).

If some of the parsing gets done in subs it is much easier to pass a reference to a context hash around than it is to pass a whole bundle of individual flag variables.

A down side with the hash is that you can end up with some subtle bugs if you don't spell the key names consistently! At least use strict keeps tabs on that using the flags approach.