Of course it's not ideal. But no model can be ideal. The Lanterne Rouge guys for example in the past have lied and invented ''headwind'' for certain climbs to increase the values of performances.
LR are just influencers ragebaiting. WattsToWin are unbiased, but they just oversimplify a model mainly I think because they want to automate as much as possible cause try to get about every race in it, so I wouldn't be surprised if they're just datascraping.
And even if the model is as fair as possible, you get all these race specific elements like especially the pacing of a climb which is hugely important, and I'm more positive than ever that the fastest way to do a climb is to just blast off from the bottom like Yates did (not like Carapaz/Del Toro did tbf), and when you start comparing to other climbs that were raced negatively where attacks happen with 3km to go on a 15km climb it's never gonna compare.
I think it's probably still more useful to look at timegaps to other riders in the same stage, and compared that to stages with similar climbs and similar tactics. And if you want to use Watts2Win index, I would much rather look at the difference to other riders rather than the absolute value.
Yesterday would be fairly comparable to climbs like Moncalvillo 2024, Ancares 2024, Verbier 2009, Ax3 Domaines 2013 for example. Ofcourse gaps are bigger, becaues Pogacar is better and more dominant, but those are all 20 minute climbs where they blasted the base and where gaps got very big relative to the stage difficulty as a result.
Pogacar did +12 on Vingegaard (+16 on Lipowitz as the fastest mortal)
Roglic did +8 on Gaudu on Moncalvillo
Roglic did +5 on Landa on Ancares
Froome did +8 to Richie Porte and (+11 to Valverde, the first non domestique)
Contador's Verbier climb doens't have any index for other riders, but presumably it would be veyr similar to Froome's.