Glycogen went up and the 100 kilometre time did not move

Why they ran it

The authors set out to test carbohydrate loading against a performance task designed to resemble the demands of competitive road racing rather than a laboratory ride to exhaustion, with sprints inside it.

Two features were built in deliberately: a placebo controlled moderate carbohydrate comparison diet, and carbohydrate taken during the ride itself to keep availability high, which is what a racer would actually do.

Drawn from the paper's introduction and stated aims.

Carbohydrate loading at 9 grams per kilogram per day for three days significantly raised muscle glycogen compared with the placebo controlled moderate diet at 6 grams per kilogram per day: 572 against 485 millimoles per kilogram of dry weight, roughly 18 percent more.

Time to complete the 100 kilometre trial was 147.5 minutes after loading and 149.1 minutes after the moderate diet. The difference did not approach significance. Mean power output was 259 against 253 watts, also not significant.

Total muscle glycogen used across the trial did not differ between the two conditions, despite the loading condition starting with more of it.

The numbers

Muscle glycogen, loading vs placebo diet	572 +/- 107 vs 485 +/- 128 mmol/kg dw, p<0.05
Time to complete 100 km	147.5 +/- 10.0 vs 149.1 +/- 11.0 min, p=0.4
Mean power output	259 +/- 40 vs 253 +/- 40 W, p=0.4
Carbohydrate, loading vs comparison diet	9 vs 6 g/kg/day
Carbohydrate taken during the ride	1 g/kg/h, in both conditions
Total muscle glycogen used	no difference between conditions

Why this might happen

Open question The authors do not claim to know why.

The authors offer two explanations and decline to choose between them. The first is that carbohydrate taken during the ride prevented blood glucose from falling at all, which would offset any disadvantage of starting with less glycogen in the muscle and the liver.

The second is blunter, and it is their own sentence: part of the benefit reported for carbohydrate loading in earlier work could have resulted from a placebo effect.

Drawn from the paper's discussion as reproduced in the abstract.

What this does not show

Seven riders cannot rule out the size of effect that decides races. The loading condition was numerically faster, by about a minute and a half across two and a half hours, or roughly one percent. That is exactly the size of difference a trial with seven riders cannot see, and it is the size a racer would care about most.
Everyone ate during the ride. That is the point of the design and also its boundary. It says nothing about an event where the athlete cannot eat, which is where several of the positive loading results sit.
One event, one duration, one sex. A 100 kilometre time trial of about two and a half hours in seven well trained men. A longer event, a hotter one, or a different sport is a different question.
A significant glycogen difference is not a large one. The two conditions differed by about 18 percent in muscle glycogen, and both were well within the range this literature calls loaded. This is not a comparison of a full muscle against an empty one.

Where this leaves us

This trial separates two claims that usually travel together. Loading reliably raises muscle glycogen, and raising muscle glycogen reliably raises performance. The first survived here and the second did not, at least when the rider is eating during the event. It refines rather than overturns the case for loading, because the trial is small and the loading condition was numerically faster.

Seven well trained male cyclists over about two and a half hours, fuelled throughout at one gram of carbohydrate per kilogram per hour.

The same design run unfed, which would isolate whether in race carbohydrate is what erased the difference, at a sample size large enough to detect one or two percent.

Caveats worth holding

Seven riders, all male, all well trained.
Dry weight units.
The comparison diet was 6 grams per kilogram per day, which is a normal athletic intake rather than a low one. This is loading against ordinary eating, not loading against depletion.
The 100 kilometre trial lasted about 147 minutes, inside the 90 to 240 minute window where loading is usually claimed to help.