The more I check it, the more confident I get despite my initial disbelief. Yes, they improved by more than 5 % overnight. Run the numbers, check different sources. It checks out.
Ok went to Pog's Strava segment for Beille and did a calculation for the final 5 km of that segment, just after Pog attacked Vinge. I get an average power of 6.81 W/kg for that segment. This is assuming no wind and no draft. For the entire climb, you get 6.57 - again no draft effect considered (which would lower the value) but also no slope&pace irregularities (which would make it higher)...
Of course I don't know many of the exact parameters, so here's the python code to play along if anyone wants to (add wind effect etc..):
import math
# simulation parameters
g = 9.8067
crr = 0.0023
CdA = 0.24
Rho = 1.11
drivetrain_efficiency = 0.98
# route parameters
distance = 5000
altitude_diff = 359
time_delta = 11*60 + 32
# rider parameters
rider_weight = 65
bike_weight = 7.8
other_stuff_weight = 0.5
gradient_percent = altitude_diff / distance * 100
speed_avg = distance / time_delta
system_weight = rider_weight + bike_weight + other_stuff_weight
F_g = g * math.sin(math.atan(gradient_percent / 100)) * system_weight
F_rr = g * math.cos(math.atan(gradient_percent / 100)) * system_weight * crr
F_wr = 0.5 * CdA * Rho * speed_avg * speed_avg
P = (F_g + F_rr + F_wr) * speed_avg / drivetrain_efficiency
p_per_kilo = P / rider_weight
print(p_per_kilo)
P.S. If there's any python coder out there, I know the code is terrible, I only use python for fooling around...