This is really the crux of our differences. You are correct that I believe these assumptions to be extremely optimistic, just as you found my 50/50 assessments to be extremely pessimistic. I have tried to base my assumptions on objective information as much as possible. Independent polling in all four of these states are well within the margin of error -- Thune typically has a 2-4 point lead in SD; Carnahan has a bigger lead than that in the majority of MO polls; MN appears to be dead even, as does NJ. A recent AR poll shows the GOP incumbent down by 8 points. These are the reasons I consider these races to be toss-ups, and why I cannot in good conscience consider the GOP to be a 90% favorite in SD, or an 80% favorite in Missouri. True, polls can be wrong. But not to the extent of 50/50 vs. 90/10. For comparison, the Political Oddsmaker puts only 3 Senate races at 90% or greater odds: Cochran in MS (95/5), Warner in VA (90/10) and Rockefeller in WV (91/9). Even John Kerry in anti-Republican MA (88/12) doesn't rate that high!
All polls I have seen by R and neutral sources say Hutchinson is ahead in Ark.
I do not think I have seen any poll at all showing Wellstone ahead since say June 1, maybe April 1.
As to Torricelli, of course, that race is just now coming into play and being admitted as very close or even leaning R. That is why I have my "all others gee-whiz" item; it seems there is always some state that had been thought secure, where just in the last few days it collapses for the D and the R gets it.
Another question is whether you see any Republican hope at all in my "little 5:"-- IA, GA, MT, LA, and IL. I just see very large imponderables not yet assessed, the campaigns are not fought yet, new faces not yet known, old fogies not yet dissed. Indictments not handed down, whatever.
BTW even using your numbers, I still get an expected R gain around 1.23 +/-, where only 0.5 or better would switch the Senate.