Very frequently on Ancestry search results, records that match the search specification on more data points will be sorted FAR below those that match the search specification in fewer fields.
This happens to me multiple times every day. I've just come to expect that one should always read several screens of search results rather than concluding that the presence of obvious mismatches on the first screen mean than the later records are likely to be worse fits.
For example, this morning I did a search in the Kentucky Death Records database for "Nola Rob*, died 1916 +/- 1 year in Grant County, Kentucky, default matching (i.e., proximity prioritized higher but non-exact matching allowed). I used the wild card in the surname to accommodate spelling variations.
The top match (and the only one in the upper section) was for a John W Roby, died 20 Nov 1917 in Kenton Co. Note that the county is only an adjacent one.
In the other results section, the highest result was for a Mason Roby died 15 Aug 1913 in Scott Co. Note that this is 2 years outside the date range as well as having only an adjacency match in the county. (Barely adjacent at that - Grant and Scott are like two diamonds balanced on their tips.)
The correct death certificate turned out to be entered as Noley Robison, died 11 Sep 1916 in Grant Co. It sorted in position 21. Sorted above it were 16 other hits that were NOT in Grant county, some of them not even in adjacent counties, and 9 of which were outside the 1915-1917 date range as well. Even treating the given name as getting zero match points, BOTH the date AND the county are better matches on the "Noley" record than for all those other records. (And I would argue that default matching ought to include brownie points given for phonetic similarity in given names, as well. If taking that into account, Noley's record should have been given top billing. Even without it, it should have been no lower than position 6.)
I'm attaching a screen shot, with Noley's record underlined, and a green sidebar marking the 6 records that look to me like they should have been sorted at the top (the only ones that have an exact match in both date and location.)
This problem is endemic. I could literally make a dozen or so such screen shots every single day on actual searches I've done.
Can anybody at Ancestry explain what the default sorting rules are, and why they seem so unintuitive? I do approve of including all those other approximate matches in the results, because the "right" record often varies from the search terms, but it seems to me that the sort order ought to correspond in a reasonable way to the number and quality of matched fields.