I'll try and be descriptive without going too deep into the somewhat poorly researched outdated science,
Classically Utilization, as described by Tinseth for example, is purely a function of the gravity*1 and time*2
https://realbeer.com/hops/research.html
*1 - Gravity Tinseth specifically generalized this to only be the midpoint boil gravity, this varies by interpretation from software to software since it wasn't consistently or properly documented or described in his publications
*2 - Time - Specifically the individual time that each hop is boiled, but this does not take into account how the pH changes over time as the wort boils regardless of the timing of the individual hop addition.
Modern challenges to these out of date formulas:
1) Doesn't account for modern hop additions not occurring during a boil. Classically, all the IBU formulas predict a 0 IBU addition to a flameout addition, or whirlpool, or hopstand. Let alone dry hopping. Thus we need to do some kind of fudge-factor for each non-boiling step, which should be dependent on the duration of the step after the boil ends, as well as the temperature (curve)
2) All the old research was done in the range of 0IBU up to around 40 IBU. Thus didn't take into account the solubility limits and diminishing returns as you approach a "soft cap" around 70 (measured) ibus, to borrow videogame terminology. As you get closer to that soft cap, the utilization SHOULD decline rapidly, but doesn't really do so.
3) All of the research was done with leaf hops, not pellets. Pellets generally have a much higher AA, but also break down and isomerize more quickly (think about how grain crush affects the conversion vs time function). Additionally there are a variety of hop products available nowadays from T45 pellets, T90 pellets, dried leafs, aged hops, aged pellets, cryo hops, lupilin max hops, hop extract and hop oils. etc.
We've introduced an adaptation to problem 1), just as many other software has, by attempting to apply a modification to the utilization and treating it as if it was a boil addition. At some point we will be revisiting this feature and taking a different approach, as right now we're independantly treating a boil addition as if it stops isomerizing completely as soon as the boil ends, while also calculating a meaningful IBU contribution from a whirlpool addition. In reality, both the whirlpool addition AND the boiled hop will continue to isomerize.
Additionally the kettle geometry places a part here as well, as the hop oils stick to the kettle walls. This was largely ignored and tried to be combined with the boil off factor, following the assumption that the difference in the "standard" range of brew kettles is not significant enough to be worth chasing as a separate variable. But it is known non the less to be a contributing factor.
