I don't like that, bodatrinadh. It'll "work", but since the datasets are supposed to be in sequence, the problem is pointing to at least one error somewhere else. If you just "get around" the problem, then you've missed the opportunity to pick up the error(s) which might be a symptom of something worse, which won't then be found for a month, which has by that time screwed up the system magnificently.
Plus, if the data is supposed to be sorted, sorting it again is a waste. That is what MERGE is for, for operating on datasets which are already in the sequence you want.