So, what design do you achieve now?

As far as I understand, you

• select randomly and also display randomly 11 out of your 33 "names" groups (each containing three questions)
• select randomly and also display randomly x out of your y "behaviour" groups (each containing three questions)
• select randomly something of your "graph" groups (6 positive vs. 6 negative  or 6 graphs randomly positive or negative?)

Well, this way you can't randomize the groups "names", behaviour", "graph".
If you really think it to be necessary, you may create 7 surveys
6 surveys that display these three "groups of groups" in one of the 6 possible orders.
1 survey that only creates a random number  from 1 to 6 and in the end-url links to the matching survey.

As usual:
If there are still questions, show what you did by lss export.

Joffm

LimeSurvey Cloud version: 6.4.4
Survey theme/template: fruity twentyThree
============
Hey, Joffm,
I've read your discussion and tried to understand the basics of randomisation.
You gave me an advice in another topic (about the mismatch). I followed the suggestion of tpartner and removed group randomization. And once I did that, the survey ran correctly.
However, the randomization and the conditioning of question. You proposed not to use randomization function built-in Lime Survey and write a code.
So, I am not familiar with Java Script. I ran the survey you've wrote for AideenW and found that I have the same type of question he had: Q1 (Tamms) -> Q2 (TammsReminder) + Q3 (TammsJa). 
I linked the necessary question to the target ones but they appear only occasionally. Despite the logic is ok as well as syntax, the survey doesn't work correctly. Is it the problem of condition? Do I have to write a condition manually to make cascade questions emerge?

I saw you stated the "relevance" to the questions. Do I have to apply this option to every linked question within the group?

Thank you for advance!