If this is your only use case, I think the simplest solution is to duplicate the loop/merge, and if they answer yes to Q5, send them to the duplicate, coded as if they had selected the inverse of their Q1 answer.

nroot Thank you for your answer!
I have thought about this solution but it is not the only case

https://community.qualtrics.com/XMcommunity/discussion/comment/48167#Comment_48167A solution I've used that is more general purpose but requires more back-end coding is to abandon the idea of one-row-one-respondent, use an ID embedded var to link together responses, and use embedded variables to control the loop. So, you have a separate survey for your loop/merge, and depending on Q1 response they get sent to the survey with their product choices flagged in embedded variables . Then, if they indicate that they want to do it for the rest, you send them to the loop/merge survey again as a "new" participant, but with different product choices flagged. Then in your analysis, you know that it is the same participant because you had an embedded var with their ID.
This requires you to merge multiple responses during your analysis pipeline, but this is often much easier than doing some sort of complicated Qualtrics routine.

@nroot thank you! I will try this option!