The model put forward by Mader and Heck made an assumption that when linearly increasing exercise intensity, ADP concentration increases at a predictable and obligatory rate which would activate glycolytic throughput in kind. This is the first of three incorrect assumptions because it ignores training status, as countless studies (some available at the time) show that endurance training improves markers of mitochondrial enzyme concentration and oxidative ATP production, which in turn reduces the rate of ADP and AMP formation with better preserved muscle glycogen while improving fat oxidation. Or as The Cog says, glycolysis is sucked, it doesn’t “blow”, meaning it’s not made at an obligatory rate. The second incorrect assumption is that lactate/pyruvate is shuttled into the Krebs cycle first and preferentially, meaning fatty acids and ketone bodies are only used due to insufficient lactate formation. Again, we can see this is not true because diet can manipulate substrate availability and relative oxidation rates without a change in mitochondrial mass or enzyme density due to training status. A tangential assumption (let’s call it 2a) is that increasing vo2max is one of two ways to improve lactate oxidation, however the model leaves out the fact that improving training status means greater lactate oxidation even without an increase in vo2max, which violates the model’s assumption that lactate oxidation rate is the primary determinant of FTP. Further reading (predating the Mader & Heck paper by several years): https://journals.physiology.org/doi/pdf/10.1152/jappl.1984.56.4.831
The theoretical backing of the model is basically dead at this point, but now we get to the kicker and the answer to your question.
The reason all this matters is the third incorrect assumption, and is the second way to theoretically improve FTP. If glycolysis were activated at an obligatory rate, then it might be possible to raise or lower this rate to increase/decrease the production of lactate by modulating (via training) the fastest rate at which someone can make lactate (which also adds large error to this measurement since we are measuring blood lactate and making assumptions about equilibration between compartments, hydration, etc). This last assumption is incorrect because the people who make lactate the fastest have the best sprint because lactate production is a reaction to the ATP usage determined by such high work rates. You could improve your sprint power with better technique and change no other characteristics about your muscles or training status, and in theory you would have increased VLamax but in reality you haven’t as you just increased rate of ATP demand which is too rapid to be met by oxidative phosphorylation, or as Cog might say, the rate at which glycolysis is sucked. The model would say that if your FTP hasn’t changed but your sprint power and lactate formation goes up, your VO2max would necessarily have to have improved and this is measurably not true. Now we get to the last and probably most incorrect thing that’s kinda in the model but mostly in the training methodologies put forth by the model proponents, which is that whether or not we’re oxidizing lactate is what determines FTP. Lactate formation is not really a causal factor here, it’s a byproduct of whether or not energy state can be propped up via oxidative phosphorylation. So not only are the model’s most fundamental underlying assumptions all incorrect, the implications for training are generally also incorrect.
The above figures are from the 1988 Mader & Heck paper and probably the easiest visual aid in understanding the assumptions in the model.