A “puzzle plot” is, as the phrase suggests, merely a plot which is in its essence a puzzle. As such, the term has a wider scope than “impossible crime story,” which in turn has a wider scope than “locked room” mystery. Perhaps I should say that the “locked room mystery” is a subset of the “impossible crime story” which is in turn a subset of “the puzzle plot.” A few examples may help make things clearer:
Conan Doyle’s “The Speckled Band” is a locked room mystery. As such, it is also an impossible crime story and a puzzle plot.
Ellery Queen’s “The Lamp of God,” which involves the stunning vanish of an entire house, is not a locked room mystery (there is no locked room involved), but it is an impossible crime story and thus also a puzzle plot (there are many examples of this type of story… Carr’s “The Curse of the Bronze Lamp” is another that comes to mind, though the word “Lamp” in the title is certainly not a prerequisite!).
Agatha Christie’s “Five Little Pigs” (as with most of her novels) is neither a locked room mystery nor an impossible crime story, but it certainly is a puzzle plot.
The term puzzle plot is used nearly interchangeably with several other terms, though there are important distinctions. It is usually a puzzle plot that is meant when one refers to a “Golden Age detective story,” but not all puzzle plots were written within the historical period referred to as detective fiction’s “Golden Age,” nor is a detective figure necessary for a puzzle plot (Christie’s “And Then There Were None” certainly qualifies as a puzzle plot, and it lacks such a detective character [yes, I know about William Henry Blore… but he quite obviously doesn’t count!]). “Whodunit” is another term that is frequently used to refer to puzzle plots, but is actually another subset; many puzzle plots are not concerned with the question of culprit identity, instead centering on questions of “how,” “why,” or even “where” (e.g. Poe’s “Purloined Letter,” where the identity of the culprit is known from the beginning).
Again, the boundaries of classification are sometimes fuzzy and in constant dispute (of course, I’ve tried to make it easy on myself by choosing examples that are fairly clear-cut). Did that help? I hope so.