I will grant you that I am no polling rocket scientist, but a poll of 1000 people across 50 congressional districts?  If spread evenly, that means 20 respondents per district, which seems like a very shallow sample from which to predict the vote of the hundreds of thousands of persons in that district.  Also, in looking at the ages of the respondents, over 65% were over the age of 40.  I do not think this represents the American electorate well, but perhaps it does.  I would like to see better data.  Finally, 52% were women, which again seems too high in respect of likely voters.  Again, I would like to see confirmatory data.  Overhanging this all were concerns about "the war", which certainly factored in concerns over the Israeli conflict in Lebanon.  This could be a non-issue by November.