In the PDF you're using, there are multiple (but not all) pages with colorful (e.g., green and orange) boxes containing text. You want the AI assistant to extract the content from these boxes, but it's important that they remain separated (in the PDF, this function is handled by the color).
Possible solution:
When you upload the PDF, you could, for example, prompt:
“Create a deck with the content of the uploaded PDF. Use only the text between headings x and y. Use only the content in the green boxes. Create at least 10 flashcards.”
Then you make another prompt; this time, to turn the content of the orange boxes into a deck. To create practice questions, you can click on the spark icon next to the deck after making the content cards.