When I ran the comparison of the original versus the most recent versions of the sequence comparisons, I saved the results of the Seq1 and Seq2 tables. While running some tests on the result tables, I discovered a mistake:



The four counts should all be equal, and because they are not, we know that there is a mistake in at least one of them. Because the version 1 results are not even consistent between one another, I have to think that the problem is with the original version.

This query:

SELECT
Seq1_Line = ISNULL(LTRIM(STR(S1.CodeLineNum)), '')
,ISNULL(S1.CodeLineTxt, '')
,ISNULL(S2.CodeLineTxt, '')
,Seq2_Line = ISNULL(LTRIM(STR(S2.CodeLineNum)), '')
,OrderBy = ISNULL(S1.CodeLineNum, S2.CodeLineNum)
FROM Seq1 S1
FULL OUTER JOIN Seq2 S2
ON S1.CodeLineNum = S2.MatchLineNum
ORDER BY OrderBy

will display the matches between the two sequences in a "Windiff"-fashion, and running it on the results of the most recent matching algorithm shows that that one worked correctly.