
He doesn't need to pay attention to the bag to having had missed about 1 ft above the shoulder.
He knows what he saw:
Mr. FRAZIER - Well, I say, like I say now, now I couldn't see much of the bag from him walking in front of me. Now he could have had some of it sticking out in front of his hands because I didn't see it from the front, The only time I did see it was from the back, just a little strip running down from your arm and so therefore, like that, I say, I know that the bag wouldn't be that long.

It is truly remarkable that you put 100% faith in Frazier perfectly remembering such a mundane detail and then turn around and treat the fiber evidence as if it is not probative at all. This is a perfect example of what I said earlier about you being really, really bad at weighing evidence. This is exactly the kind of detail an eyewitness would get wrong. People just aren't that observant about mundane details such as this. Our minds are not equipped with a DVR. We just don't take in every detail, especially something that at the time wouldn't have seemed the least bit important. Then you turnaround and act as if the matching fiber evidence isn't probative at all. It would be a truly remarkable coincidence if the fibers that were in the bag matched the fibers in Oswald's blanket if they had come from another source. Ditto for the fibers on the butt plate of the rifle that matched his shirt. To argue for Oswald's innocence, it is necessary to believe the least likely explanation for dozens of pieces of evidence and disregard the most likely one.
If you don't believe me regarding how unreliable eyewitness testimony is, maybe you will believe people have actually studied this subject:
https://pmc.ncbi.nlm.nih.gov/articles/PMC9451081/https://legalclarity.org/what-percent-of-eyewitness-testimony-is-accurate/https://www.sciencedirect.com/science/article/pii/S1364661325000270Here are some quotes from the above articles which pretty much says all that needs to be said about the reliability of witnesses:
"These numbers mean that eyewitness identification is useful but nowhere near the gold standard that jurors tend to treat it as."
"Work by Elizabeth Loftus and her colleagues in the 1970s showed that eyewitness memory, like any type of forensic evidence, can be contaminated [1]. For example, a true memory of a stop sign can be replaced by a false memory of a yield sign based on nothing more than a passing suggestion [2]. Later, in the 1990s, it became clear that it is even possible for someone to acquire detailed and emotional false memories of traumatic events that never happened"
One of the factors in witness reliability is how much time passes between the event and the testimony. Frazier was interviewed by the WC on July 23, 1964, a full 8 months after the assassination. The pictures you posted are not dated but based on Frazier's gray hair and the fact he was only 19 at the time of the assassination, I would gladly wager they were taken at least 50 years after the assassination. That gives a lot of time for memories to fade and false memories to take hold.
I have served on two juries in criminal trials. We convicted one person without relying on any eyewitness testimony at all, using only the forensic evidence to find guilt beyond a reasonable doubt. The other was a murder trial in which the accused admit to the killing but claimed self defense. There was some eyewitness testimony but it wasn't crucial to our finding of guilt. We simply didn't believe the killer's claim of self defense.
If I again were to find myself on a jury in a criminal trial and the prosecution's case hinged on eyewitness testimony, I would vote to acquit unless there was compelling evidence to corroborate the eyewitness account. I simply do not have enough faith in eyewitness accounts to remove reasonable doubt without corroboration.
But you go on treating Frazier's memory as if it is gospel. It will continue to prevent you from knowing the truth of the JFKA.
You are certainly among the people who treat an eyewitness account as a gold standard of evidence.