How granular can the question draw get?

For this specific qualification, we want our test to have three items from each chapter. There are 30 items available for each chapter, three per ‘section’ within a chapter, and ideally we would have a a random selection whereby one item is chosen from each of three of the ten sections (or pools). This model would be replicated across each of the chapters, with three items randomly selected per chapter across the qualification. 

To summarize:

-10 chapters

-10 sections within a chapter

-3 questions per section 

-Draw 3 total questions per chapter using 3 different sections

 So the question is, can we do this in Articulate? And if not, what are the limits to the level of granularity in how we define the pools and selections from them?