Sure, until you get the current questionnaire all entered, leave it how you have it. However, in the long run, you need to normalize the tables to get sensible information out of them.
Look at it this way. On Questionaire # 73, Section E, word number one is "Apple". However, on Questionnaire #75, "Apple" is word number two in section D.
The question (or prompt) is "Apple". You store "Apple" in the question list, for example as prompt number #137, and in the junction table you store the information that allows
Code:
Junction table example
QuestionnaireID 73 75
SectionNumber E D
QuestionNumber 1 2
QuestionID 137 137
Storing the information this way not only saves space, it also allows you to compare the results on the same prompt from different surveys, regardless of where they happened to be on the particular survey. If you don't normalize the database, that analysis will be nearly impossible to achieve. With a normalized database, it's almost trivial.