ABC claims that LOST finally tamed CBS to seize the hotly contested 9pm time slot for competitor Criminal Minds. In ABC's press release, the victory was not only solid but had a little ring of finality to it:

“Lost” won its time period and ranked as the No. 1 TV program of the night in Adults 18-49 for the 5th straight week (on each of its telecasts this season), defeating second-place CBS’ “Criminal Minds” in the slot by 35% (6.6/16 vs. 4.9/12). However, Zap2it has an assessment of the Wednesday night battle that would warm CBS's heart: 'CBS Tells ABC to get LOST'. By their numbers both networks were up, but CBS again dominated the target demo as well as besting LOST, for the first time, in total number of viewers. So who has the right data?
The answer could be: neither. Both assessments are the results of early projections or 'fast' numbers. In Zap2it's case, they are relying on Nielsen's early data which is derived mostly from Tivo scheduling. This data usually amounts for less than half of the reporting families and can pack an error margin that would make ABC's numbers realistic. Networks, naturally, have their own data collection mechanisms and are often better at predicting the final Nielsen numbers than Nielsen themselves. Why? The networks are able to trend through non-nielsen sources and volunteer media tracking organizations which mirror the Nielsen model, Nielsen does not go to these lengths to process their fast data. They merely pose the percentages against their contributing families to come up with a hypothetical number based on the assumption that all families will behave like those tivo users. While it does happen from time to time, it is pretty uncommon for a network to claim a win this huge without a great degree of certainty. With the slot this highly contested, you can be sure both networks are waiting for the final numbers with bated breath.