My main argument is that wait times are fluid regardless of fastpass (and posted wait times are inaccurate in general), and should not be used as a measurement with regards to late FP use.
You also get the exact same effect if a large group showed up with their Fastpasses at the same time, but they were all within their window - some at the end of their window, some in the middle and some at the beginning.
The way my argument functions is that everyone using the standby line is going to enter that line at a certain time of day. We have to leave out whether they decide not to because of the wait time indicator, which as already mentioned is inaccurate and we cannot predict what the "breaking point" is for those people, so we presume they will ride regardless as was the case before FP.
Here is something of a rewrite, because for unexplained reasons I'm feeling scientific...somebody needs to stop me so I can actually get some work done...
There are a few assumptions to make this case fairly simple, such as no maintenance issues affecting the load rate that day, the standby line is never empty, etc.
Let's call that person S.
Now, when they actually get to BOARD the ride will depend on the number of people in front of them in the Standby line, and the number of people who use their Fastpasses, between the time they enter the Standby line and the time they board. If there was no Fastpasses, then S gets to board at a known time, say N. With Fastpass, they will get to board sooner because some number of guests are deferring entering the standby line by getting a Fastpass for a time later than N, so call that new time T. Early on, N and T are close to each other but then separate until a theoretical maximum, and then towards the end of the day they shrink back together again.
If someone did not use their Fastpass (call them person F), which was eligible any time before S's board time, until after S gets to the boarding area, then S gets to board sooner than T (call the difference B for "bonus"). For X number of Fs in this case, then S gets to board at T - (X * B).
If any F DOES use their Fastpass before S boards, then S does not get their "bonus", and the time moves back closer to T. This does not actually change if they used their Fastpass on time or late, or whether they did so hours before or walk up just before S gets on. S will get to board no later than T.
The problem in the standby line is that they PERCEIVE that they had to wait longer because from their frame of reference, they were going to get to board at T - (X * B) (and which may have been corroborated by the wait time indicator), but that changed because one or more Fs showed up. The variance obviously increases with more Fs using their Fastpass while S is in line.
Now if everyone used their Fastpass at the first possible minute, things would be fairly constant. The wait time indicator would be fairly accurate (at least with regards to Fastpass use, but a single Tour Group can break that). But not everyone uses their Fastpasses that way...after all, you have a whole hour, right? So some come first minute, some within 15 minutes, and some within the hour. They come in clumps. So there are perturbations to that bonus time that S hopes to experience. But they have not affected T.
So, what if F comes later than their hour window? Nothing changes with regards to T. In fact, even more Ss experience a bonus while they are late. But it does mean that T - (X*B) is greater for S, and therefore their perception of a delay can be that much greater if F shows up.
Now, where the REAL effect is if a late Fastpass user (L) shows up before other Fastpass users who are within their window (F). Assuming the Fastpass line empties periodically (and if all is running well, I believe it should), then L has pushed back all the Fs behind them by one person. That could mean a delay of 0 (they still get loaded on the same vehicle/theater/etc.), seconds (on a continuous load attraction), or a cycle (they got pushed to the next vehicle/theater/etc. because it is now full). But even that effect goes away once the FP line empties again (i.e. Fs coming after the FP line empties the next time are not delayed at all). Of course, the more Ls there are, the more Fs that are affected and the longer the delay is.
The biggest concern is at the end of the park day - where if everyone elected to hold their FPs until the last possible moment and then crash the FP line, then there would be great delays in the FP line, and the standby line could come to a standstill unless the CM is applying the ratio. But this doesn't generally happen to that degree. MOST people use their FPs as soon as they reasonably can.
ARGH!!!!! Why do I keep doing this to myself?!?

I hope ANY of that made sense.
The 80:20 rule seems a bit much. That implies that they are giving out Fastpasses at a rate equivalent to 80% of the load rate, which seems awfully high. Not impossible, but it just seems that way from observation. It also implies that 80% of guests must be using Fastpass, and I REALLY don't see that. Perhaps it is a maximum ratio to deal with the perturbations I've mentioned, along with other reasons (like maintenance issues).