I read some previous post about setting flexible factorial and contrasts in flexible factorial but I've got a question I did not saw any response about it.

I'm trying to analyze a 2x2 repeated measures design (no group factor, 2 conditions: A and B).
So I built a Flex Fact with : 

Factor 1 : subject

Factor 2 : Condition A

Factor 3 : Condition B

And follow the Glascher's tutorial about contrasts in Flex Fact (Maybe I will use GLM Flex, as it seems powerful!)

However, I also read some fMRI papers with a similar design and I was very surprised to see that some Flex Fact were built with one "experimental" factor.

For example, with a 2x3 design, the authors used a Flex Fact with one factor with 6 level: A1B1, A1B2, A1B3, A2B1, A2B2, A2B3. I guess that there is a (theoretical or statistical) reason to make this choice, but I really don't catch it. 

Any idea ? 

Alexandre