Unfortunately I have no references, but I'm pretty sure that historically even really pedantic people were less precise about these things.
The root of this is a different interpretation of sentences using the word "all", and issues related to this caused disagreement even among philosophers/mathematicians into the 19th century, never mind everyday usage. Consider the difference between William Hamilton's "quantification of the predicate" and the logic of de Morgan and Boole.
The construction "all cars are not made equal" can mean the negation of "all cars are made equal" if you interpret "all cars are made equal" in a certain way to begin with. Parse it as (all cars) are (made equal [to each other]), then the negation of that is (all cars) are not (made equal [to each other]).
This is perfectly respectable in English, albeit potentially ambiguous. It's unfortunately different from what you think it should mean: every car is not equal [to any other car].
With the lemons it's even more reliant on that group meaning of "all". "All fruits are lemons" would be taken to mean the that group of "all fruits" has the property, collectively, of being lemons. So far so good, but the symbolic logic version "every fruit, individually, is a lemon" is logically equivalent but structured differently. Then the first interpretation of "all fruits are not lemons" takes the "not" to say that "all fruits", collectively, do not have the property of being lemons. It's not "all fruits" that are lemons, it's only "some fruits", so one says "ah, but all fruits are not lemons". Whereas you expect the negation to fall inside the the "all" quantifier in the modern symbolic-logic proposition, to result in "all fruits have the property of not being lemons", that is "every fruit, individually, is not a lemon". Which is false.
As the link in comments shows, another example is "All is not lost", from Paradise Lost. This means "not (all is lost)", it doesn't mean "each thing is still here". Etymology gets tougher that far back because of the relatively small volume of published work -- even if Milton is the first reference we have (and I'm not claiming it is), then that doesn't mean he started it. No lesser poets than the lads of Coldplay maintain the tradition into the present day with a song named "Everything's Not Lost". There's certainly plenty of precedent for the "logically wrong" usage, and I don't know of any reason to hold that your meanings are more correct in English.
Your preferred meanings are, I agree, more logical by modern standards of the quantifier "all". But if logicians were in charge (rather than real observed used of the English language) then "I ain't got none" would mean I have some. Which it never means. For all that it should mean that, and for all that people's parents and teachers tell them it means that, it continues not to mean that.