After all, the primary form of evidence for any given memory is the verbal recounting of detailed information. Without language, there aren't many ways to externalize a memory. Whether one is describing the experience of a memory, or reciting a coached, scripted tale ...if all the facts are correct what hope is there of disproving a dyed-in-the-wool memory?
Case in point: The drunkest I ever got was during a drinking game in college. I was black-out drunk. My memories are truncated about halfway through act one, and the next thing I honestly remember is waking up to the stench of my own vomit covered bed in my dorm room, but such an adventure it was for my roommates, and such a bardic tale it was for them to recount, over and over again, ad nauseum, that now, I too, can recite the tale, moment by moment from when I blacked out until I woke up. But, to the outsider, my recall seems accurate, and is vigorously affirmed by my roommates, and based on my ability to tell the story, a listener has no honest means of determining whether I was too drunk to remember or not.
So, while I buy into the mechanisms for how OP's claims to memories could be false in that he is perhaps only remembering the plot of the story, and not the experience itself, I'm still open to believing OP's claim to certain vivid memories.
I don't think childhood amnesia is a bottomless black pit from which nothing ever returns. Very obviously, certain aspects of learning must survive, and I think the keystones of our emotional development are indelible. Early traumatic experiences stay with people for very long periods of time. Being bullied, humiliation, loss of a pet. Those are foundational learning experiences. The detail of the bulk of our day-to-day experiences may get lost, but set pieces, scenes that perfectly encapsulate a profound learning experience are different from the tangential recall of banal routine. I think animals are wired to retain certain types of experiences, even from very young ages.
The phenomenon of memory is not necessarily constrained by hard statistical numbers. At least that is my opinion.
I'd offer up similar evidence of my own experiences, although I'm sure they'd be readily dismissed as subjective and unsubstantiated.