I've started typing my answer a few times, now...
There are a few things that make people say this.
1) Studies are small.
Most studies in ME are low-powered and viewed by research scientists as "maybe interesting; maybe nothing". To clinicians, who seem to want meta-analyses of 200 double-blind, placebo-controlled studies before recommending a glass of water in the morning, it means zilch.
Low power means a high likelihood it's chance.
2) Poor methodology & execution...
So often it is chance, and Research Group B tries to replicate a study and it doesn't work out like Research Group A said it should. Or Research Group A tries to replicate its own results in a larger cohort, and it fails.
This can be because of poor initial methodology / execution from Group A, or poor methodology / execution from Group B... or it can be that Group A genuinely found that data using good methods, but see #1: could've easily been statistical chance.
In our field we have higher incidence of poor methodology and execution.
Why?
Because there's no money, so the median grant is small. That means poorer researchers and new/young researchers (both of whom may only be able to aim for smaller grants) are going to apply in our field. That isn't to say there aren't amazing people working on ME -- there are! -- but that this isn't the median.
3) No replication at all.
Found a good theory? Seems to work out in your small group of patients? Awesome!
Spin the hell out of it in the press -- and never, ever replicate.
It's really tough to get funding for a medium-sized study in this field. And besides, what if your genius theory is proved wrong in a larger cohort? That could bring in less funding in the future...!
No, no... better to swing to some other, completely new idea and get funded for the next mini-study!
(Sorry, this is the modus operendi of at least one group...)
4) Fatigue conflated with ME
So far as I'm concerned, our real issue isn't so much that fatigue studies exist, or that studies on Fukuda or even Oxford exist, but that their results are conflated with results using any of the criteria that require PEM.
It's interesting to see what, if anything, all "chronically fatigued" people have in common. So long as you don't label your study as being about ME!
ME an absolute minefield of weird assumptions, illogical conflations. How are you supposed to know all this as a researcher? It takes a year or two to get a handle on all the strangeness in research so far, and that's if your exclusive focus is ME.
One onus against the name chronic fatigue syndrome that I haven't often heard (but wish I heard more often): it invites newcomers to the field to conflate the symptom of chronic fatigue and the disease as a whole. It's usually one of the first things you have to explain if you say the name aloud. "...but it's much more than the symptom of chronic fatigue". Ugh.
5) Dogma
I'm not sure why, but there are a lot of ride-or-die theorists in ME. They'll come up with an idea and stick to it, come hell or high water. We need open-minded research: come up with a hypothesis but don't marry it and ask us to throw rose petals at the wedding. Hypothesize, test, and if you're wrong, move on.
I just came across a paper again from the Norwegian researchers who worked on adolescent CFS using clonidine.
Their first paper on this, several years ago, started from the assumption that CFS was linked to anxiety, and so if they blocked adrenaline/noradrenaline with clonidine, their patients would improve.
They didn't. They worsened on actimeter.
The researchers (rather sensibly) concluded that the increased adrenaline and noradrenaline was adaptive, and blocking it might be a poor idea. I thought at the time that it was refreshing they could admit they'd been wrong.
Now this month they came out with a paper showing that there are two groups of pwME: low adrenaline, and high adrenaline. It appears they're still using clonidine. And they've collected yet more evidence that they're wrong: pwME with higher adrenaline did better on two subjective fatigue surveys.
Your. Data. Says. You're. Wrong.
Stahp.
Oi, this raised my BP & HR.
6) A lot of the results we've found & replicated are NOT unique.
Hand grip. NKC activity. Metabolomics.
These all mimic things we've seen in other illnesses or conditions.
2-day CPET seems to produce a pretty standard pattern... in some pwME. There's a second group who, according to one clinician who's worked on this for decades, look more like very elderly people on day one & replicate that on day 2. I had a one-day CPET where the ANS specialist said I functioned "like an 80-year-old" so this checks out. So even PEM has its variance.
______________________________
Ok that was all a little dark and a bit negative, so.
Researchers & clinicians need to listen carefully to pwME in order to be successful.
If you're in science and you're reading this, there is so much you don't know that a pwME can tell you. The fact that you're here and listening speaks well for you. Get a pwME with a science background to help you design your study. It will go SO MUCH BETTER. You will avoid SO MANY MISTAKES.
Don't do it for brownie points. Don't do it because your grant says "you gotta".
Do it because you'll produce a crappier study without us.
It's true.