Language changes constantly, and meanings often flip from one century to another. The previous meaning of "prove" doesn't obviate the modern usage inside or outside of the phrase in question or turn the clock back. That particular horse escaped the barn, long, long ago along with tens of thousands of other changes in the last few centuries. "Probable" existed before probability theory did, and it didn't refer to odds or math; it just meant "respectable."
I could follow your own suit, I suppose, and compliment you by calling your comment "smart", since that word used to mean "painful" in olde english and you prefer the old ways. But I won't. (I'll just say your comment was interesting, and not quite the full story.) I roll with the common currency: language as it is. Print greatly slowed these sorts of changes, and the internet will slow them even more, of course, as the decades and centuries roll on.
In general, I try to correct it in order to give the benefit of the doubt. Either somebody is misusing a phrase, which is a charitable interpretation, or somebody has a fundamentally flawed idea of how evidence works, which is rather uncharitable.
Basically, you can name the exceptions off the top of your head, but you could never do that with the non-exceptions because they're so numerous. Kind of like how in your parking example, you don't need to name the hours that parking is allowed, because it's a rule. It's always allowed, with some exceptions. You only need to point out the exceptions.
I would take your charitable interpretation and say you're misunderstanding the phrase. Since there are examples of this usage all over the Internet and a very long Wikipedia page detailing the various forms, I hope we can move past this wildly off-topic conversation.
In this case, the appropriate term would be "survivorship bias". Those who survive are noticed and used to generalize to the group, even if their survival is atypical of the group as a whole.