I can tell you what worked for me.
I did 3 image streams for the one question.
I debriefed each one individually. There I looked at key words, example I might have used the words "lush grass", I look at why I used the word lush because there are many other words that could have described the grass, so what does lush mean to me, extravagant, luxurious, slack, loose (note they aren't necessarily the dictionary meanings, the meaning and slang relate to me how I grew up relating to the word. Hence the code of the mind) To me lush most strongly means slack or extravagant more of a detrimental word which is why I would look at my use of it in the flow of an image stream. Because I could have used vibrant, flowing, wild, long... so here is the word lush.
Looking for those sort of hooks and cues during my debrief.
I don't normally get a full answer from a single image stream. That's why I do three.
Then I look at common, feelings, sensation, meanings (cue words)I usually get some sort of message. I wrote that message, which was a sentence or two down. It took me about 2 weeks with the high think tank before I saw my answers being clear and relevant to my questions. The key was looking for what the three image streams seemed to be saying that was the same.
So from all three image streams I might write something like "cut yourself some slack" Then look at the question. (yes I write an answer before I look at the random question I pulled for the HTT). Then I explore the answer against the question.