I have three sources I'm dealing with:  Broderbund's 1790 Census Index (CD-311) and subscriptions to the indices and page images online at Ancestry.com and Genealogy.com. 

There's a sequence of pages, 117 thru 131, from Cheshire Co., NH, that are one (roughly) alphabetical sequence, so they obviously represent the enumeration of one geographical location (the digital images are two-page spreads, so each image includes two census pages).  A large number of individuals are involved because each page has four columns of names, and there are about 75 names in each column.  The Broderbund CD shows these as "Cheshire Co., NH," without breaking them down into towns.  And how could they be indexed by town?  There's nothing on the pages to indicate the town.

These same page images, containing the single alphabetical sequence, are duplicated 31 times at Ancestry.com (!), with each set having a different town label and with individuals being indexed in different towns.  I see no conceivable way to tell which town an individual is in.

The situation is even more bizarre at Genealogy.com, where each page image is duplicated and given a different town name, even though they are clearly one alphabetical sequence, namely:

The image of pp. 117-118, which contains surnames beginning A and B, is given twice, labeled Croydon the first time and Lempster the second time.

The image of pp. 119-120, which contains surnames beginning B thru E, is given twice, labeled Dublin the first time and Charlestown the second time.

The image of pp. 121-122, which contains surnames beginning E thru H, is given twice, labeled Alstead the first time and Hinsdale the second time.

The image of pp. 123-124, which contains surnames beginning H thru L, is given twice, labeled Langdon the first time and Charlestown the second time.

The image of pp. 125-126, which contains surnames beginning L thru P, is given twice, labeled Gilsum the first time and Marlow the second time.

The image of pp. 127-128, which contains surnames beginning P thru S, is given twice, labeled Chesterfield the first time and Cornish the second time.

The image of pp. 129-130, which contains surnames beginning S thru W, is given twice, labeled  Charlestown the first time and Croydon the second time.

The image of pp. 131-132, which contains surnames beginning W thru Y, is given once and is labeled Packersfield.

And the weird thing about the index at Genealogy.com... like the one at Ancestry.com, individuals indexed in Cheshire County include the town names.

There are some stray pages for New Grantham, Plainfield, and Protectworth on another microfilm roll, but let's not complicate things further.  However, what's a little disturbing about these pages is that they are written in an entirely different hand in a different style and format, with the town name very obvious, which makes me worry that there were once such pages for the rest of the towns, which have been lost.  Was the index compiled from a different set of pages, ones that included the town names?  If so, why were those pages not digitized?  Are the images we're viewing a handwritten copy of missing pages?  Does anyone know what's going on here?

And the literal bottom line is:  It seems to me that until we understand where the town locations in the indices came from, we should ignore them.