After reading your post I realized that Galic Ho was talking about LeMond the racer, not "Le Monde" the newspaper
Anyway, the original calculations that made so much noise were from Frédéric Portoleau and can be found here
http://www.cyclismag.com/article.php?sid=5184
There is no mistake due to normalization to a 70 kg rider. Portoleau did calculate 99.5 ml/mn.kg for those 20 minutes55s.
CONTADOR PLUS FORT QUE BASSO ET ARMSTRONG À LA MONGIE
Contador réalise une performance exceptionnelle avec 490 watts de moyenne en « puissance étalon » pendant 20min55s. Il développe 445 watts en puissance réelle avec 62 kg de poids de corps soit un rapport poids puissance de 7,2 W/kg. Sans aucun doute, son plus bel exploit en montagne. Pour des ascensions relativement courtes entre 20 et 30 minutes sur le Tour de France, nous n'avons pas relevé plus de 460 watts (Basso et Armstrong à La Mongie en 2004) mais ce type de final n'est pas si fréquent sur le Tour. Nous avons plus souvent des escalades comprises entre 30 et 45 minutes. Le « record du monde » appartient à Bjarne Riis avec 480 watts pendant 34 minutes à Hautacam en 1996. La performance de Contador apparaît inférieure.
He calculates a real power (not normalized) of 445 watts for 62kg, ie 7.2 watts/kg.
Since many people were shocked, as soon as Sorensen SRM data became available
http://www.srm.de/index.php? option...09&catid=112:le-france-09-blog&Itemid=260〈=us
he used them to verify his earlier calculation, but
contrary to what i may have written elswhere this did not change his estimate.
You can find here his later assessment.
http://www.cyclismag.com/article.php?sid=5207
Not only does he give the Sorensen data, he gives à 1/25000 map with partial times of Contador on the climb, arrows showing the direction of the wind, in fact
everything one could possibly want to prove him wrong, but nobody has done that and certainly not the sportscientist source that you provide.
However, as far as I am concerned he overestimates the result by 2.5% since he makes a 2.5% transmission loss correction which should not be there as Sorensen was using a SRM device that measures the power at the crank, before losses occur.
So his result should in fact read 97 ml/mn.kg for those 21 mn or so.
Had he not made that 2.5% correction, he would have been spot on in his estimate of Sorensen power!