Method
Why flaccid readings wander
Flaccid length responds to temperature, arousal, time of day and recent activity, which is why the literature reports it with wide spread and few people should bother taking it.
A flaccid measurement taken twice on the same person, an hour apart, can differ by more than a centimetre without either reading being wrong. That is not measurement error in the usual sense. It is the actual physical state changing between readings, which is why flaccid length is the least useful figure this site's method produces.
What moves it
Several ordinary things shift flaccid length on their own, independent of anything a ruler measures. Temperature is one of the largest. Cold tightens the tissue and shortens the reading, warmth relaxes it and lengthens the reading, and the effect is large enough to move a reading by more than a centimetre without anything else changing. Temperature's effect on the flaccid reading specifically is worth its own look, but the short version is that a cold room and a warm shower can produce two different numbers from the same person on the same day.
Recent arousal or activity is another. A flaccid state reached a few minutes after arousal has subsided is not the same tissue state as one measured first thing after waking, and the two do not average to something meaningful. Time of day compounds this - flaccid length tends to run smaller in the morning and larger later, though the pattern is not consistent enough between people to be a rule anyone can rely on for their own reading.
Why this makes the number unrepeatable
The problem is not that flaccid length is hard to measure carefully. The technique is the same landmark-and-ruler method used for any length reading. The problem is that the thing being measured keeps changing between attempts for reasons that have nothing to do with technique, so three careful readings taken across a day can disagree by more than three careless readings taken back to back would from pressure alone. A number this unstable is a poor basis for anything - not a percentile, not a comparison against a previous reading, not even a comparison against itself.
This is also part of why flaccid length correlates only loosely with a person's erect length. The relationship between the two states is its own subject, and the flaccid figure's instability is a large part of why that correlation is weak rather than absent.
How studies handle it
Studies that report a flaccid figure at all try to remove as many of these variables as they can before taking the reading: a warm, private room held at a consistent temperature, a rest period before measuring rather than measuring immediately on arrival, and the same standing posture used for every other reading in the protocol. None of this eliminates the variation. It narrows it to something a study can report a standard deviation for, which is a different bar than making any single reading trustworthy. That instability at the individual level is exactly what shows up as a wide spread in the pooled numbers - how much flaccid variance actually appears in the published data puts a figure on it. What the published data actually reports includes a flaccid figure with a wider spread than the erect or stretched figures beside it, and that width is the state variability showing up in the pooled numbers rather than a flaw in the studies.
What to measure instead
If the goal is a number that means something on its own or against a study, stretched flaccid length is the better target. It is taken to a resistance point rather than a resting state, which removes most of the variables above, and it tracks erect length considerably better than resting flaccid length does. An erect reading, taken at full rigidity, is more stable again.
None of this variability shows up in a photograph, either. A picture catches one instant of state along with everything else, which is one reason an image-based assessment cannot recover a physical measurement from a picture even before scale enters into it. A photo-based rating service like Penis Rater is scoring what is visible in that one instant, not tracking how it changes, and neither is a human judge's read of a photo - Rate Penis is upfront that an opinion is not a state-controlled reading. None of that is a criticism of those services. It just means flaccid length, with all its wander, is not the kind of number any of them is trying to produce, and comparing a subjective score like the ones Rate Cock generates against a flaccid reading compares two different kinds of variability that happen to share a subject.