Memoir and autobio can be tricky. Nevertheless, I firmly believe that anything presented as fact should be fact, not reconstructed memories subject to fictionalization and finesse. Future historians may rely on your book generations from now and gain a skewed understanding of the past - OK, that's an extreme example, but a real one. Particularly with regard to quotes, there are plenty of other options - summarizing, paraphrasing, presenting inner thoughts in italics, etc. - that a writer can use instead of making stuff up. Do your due diligence - talk to others who were there, consult written records, verify as many of the details independently as you can, and that'll put you on firmer ground. YMMV - just one man's opinion.
My impression after reading several articles on this elsewhere is that the efforts you describe are unnecessary. It was a surprise to me too, but that appeared to be the consensus. They also addressed your concerns (and mine also as it happens) by saying that many authors have a hard time deciding between "fiction" and "memoir" for the reasons you state. However, the rule seems to be that a memoir is about how you remember something as opposed to ensuring that every memory is objectively perfectly accurate.
On reflection, that actually does seem reasonable to me because it would be almost impossible to write these things otherwise. In my own case for instance, the events did occur, but there is no way to go back in time and verify the details. I wish I could paraphrase conversations, but even that level of detail is not available in most cases. If, for instance, I know that I sold an antique watch to a certain person in a specific place, but that person is dead and there is no record of the transaction, how am I supposed to write the scene where the transaction takes place? This seems to be where the "it's your memory" rule seems to take over.
I just finished reading a couple of excellent books by the historian William Dalrymple. In them, it is quite obvious that he only describes things if he has reference to support them. If he reports a conversation, he states the source and reports what the source said, often quoting it. The style is wonderful to read, but works only when a good amount of documentation exists.
Getting back to the watch example, it is not fiction to say that it was sold, to whom, by whom, or where. Any dialogue I make up for the scene probably does not match whatever was actually said on the occasion, but if that was the rule then very few memoirs would contain dialogue, if any. It seems like it would be a problem only if the dialogue was meaningfully different as opposed to only being different. In that sense, "I'll buy the watch" is not meaningfully different from "I'll take it, but I think you charge too much." A meaningful difference would be if someone wrote something like "I need a new watch because I broke my old one when I escaped from prison last night." This is because the only remembered fact, selling the watch, does not include the detail about a prison escape.
If, on the other hand, the book hinges on the fact that this person claimed to have just escaped from prison and had done so, dialogue could include that information even if it was not originally delivered in the context of buying the watch.
I'm just reporting my impression of these articles btw, not trying to corrupt anyone's journalistic integrity here. My instinct was to call my book fiction, but I found that writing the query letter was very difficult because the material really isn't fiction. After reading some more on the subject, I found it was much easier to describe the work as memoir, and think it is a more accurate definition.
AP