I think I got the answer!!
In their latest data entry, next to "HD DVD" there is an asterisk, which refers to a note saying: "Includes HD DVD Combo". The sales info for weeks ending Jan 7 and Jan 14 did not include it.
Hence, it might be construed that the first two data points published did not consider combos as HD DVDs, which made HD DVD look worse than it should. Probably some industry agent contacted Nielsen to correct their data.
More circumstantial evidence is comparing the chart of best-selling titles on HMM and davisdvd (thanks to darin for pointing this out). Superman Returns appears on davisdvd's list but not on HMM.
Short story: the first two weeks are useless.
Interesting idea, but somehow it does not sound right.
the SI changed way too much and the YTD did not. If the data was not right and there was a correction, I would guess there would have been something simillar for the YTD.
The issue is that it is obvious that Dec was when the surge happened (maybe late Nov).
Back in Aug/Sep HD DVD was outselling BD 3:1 (roughly).
From the HD DVDprg we get 175k+ sold, other indications show that around 100k HD DVD add-ons sold last year as well and then we have the 2G that only showed up in Dec (or was it late Nov).
Let's call p the amount of time it took to sell 10 BDs in Aug 2006
10 BDs then
3:1 -> 30 HD DVDs then
4:1 (a guess based on number of players then compared to Jan) -> 120HD DVDs now
2:1 (sales for BD vs HD DVD)-> 240 BDs now
so we get 360:40 or 9:1
I am guessing 9:1 is also extremely conservative (i.e. most likely bigger).
I can't see how it could influence SI so much and not YTD