And I certainly believe that back in the olden days (77...), films were
spliced from many different sources before being sent out for copying,
for release to theaters.
Well, that's just not true.
Here's how it works:
The final locked cut is finished using workprint, and a cut list is made. The cut list is a text list that has, in feet+frames, every shot in the locked cut, in the order they exist on the original camera reels. A second list (sometimes this information is included in the same list) has all the same shots in the order they exist in the cut, in feet+frames for each reel of the locked cut, so they know which order to put things in.
The negative cutter then takes the original camera negatives, cuts them up according to the cut list, and reassembles them to match the final locked workprint - this is called "conforming."
Sometimes this was done using A/B reels - in this process, shot 1 is on reel A, while that same portion on reel B is blank leader. Shot 2 is on reel B, while that same portion on reel A is blank leader. Any time there is a dissolve, the first shot of the dissolve is on one reel, overlapping the second shot on the other reel - the overlap is where the dissolve will happen.
Reel A is printed onto a third piece of film, then reel B is printed onto the same piece of film. This is the interpositive (IP), as printing a negative to a new piece of film creates a positive. Any dissolves are done by programming the machine to turn the light down for the duration of the dissolve on the first shot, and turning the light up for the duration of the dissolve on the second shot.
This method gives you a seamless IP that won't have splice marks of any kind, and was often used for anamorphic films.
HOWEVER, as splice marks are clearly visible on pre-SE transfers of Star Wars, the conform must have been done using a single-reel method.
This method is basically just splicing the negative together to conform to the locked workprint. Any dissolves have to be optically printed, making a new negative that is two generations away from the original quality (as you have to make an IP first, then an internegative from that - the internegative is what gets spliced in with the rest of the o-neg). The same thing would have been done for the wipes and any composite effects (lightsabers, bluescreen shots, et cetera).
What does all this have to do with the topic?
Well, the audio mix is done separately, using a copy of the workprint. This is later married to the IP created during conforming, which creates the master negative that release prints are struck from.
This is where all of this becomes relevant to the topic.
So, if there were prints that somehow had extra shots (like the rumored double-grappling-hook-throw, or the rumored lack of the mouse droid shot), it would completely throw off the sound mix sync, and require that a completely separate conform would have to have been done in order to accomodate these extra shots.
Basically, it's a fuckton of extra work that would cost too much money to bother with. It's just illogical for there to have been more than one version of the film in 1977 (discounting audio mixes, which all sync to the exact same version of the film).
Even the 1981 re-release with Episode IV added wouldn't require a separate conform - it just requires the first shot to be replaced, which projectionists could easily do. It doesn't screw up the sync of the sound mix or anything, as the frame count is the same.
Basically, it just makes no sense for there to have been multiple versions of the film on its initial release. Fox certainly would never have allowed it on such a risky film.
Memory is a funny thing - it cheats, and you shouldn't trust it with something you remember from 33 years ago.