This seems to work as desired.

WITH Scores AS(

SELECT EVL.Account,

EVL.Ident1,

EVL.EvalDate,

FIL.Module_ID,

AVG(CASE WHEN ANSWER_VALUE IN (0,1,2,3) THEN ANSWER_VALUE*1.0 ELSE NULL END)*10.0 AS Score,

DENSE_RANK() OVER(ORDER BY EVL.Ident1 DESC) ranking

FROM #evalanswers AS ANS

JOIN #filter AS FIL ON ANS.Answer_ID = FIL.Answer_ID

JOIN #evaluations EVL ON EVL.Ident1 = ANS.Ident1

WHERE ANS.Answer_Value IN (-1,0,1,2,3)

AND EVL.COMPLETE = 1

GROUP BY EVL.Account,

EVL.Ident1,

EVL.EvalDate,

FIL.Module_ID

)

SELECT s2.Account,

s2.Module_ID,

MAX( s2.Ident1) AS IdTime2,

MAX( s1.Ident1) AS IdTime1,

s2.EvalDate,

MAX( s1.Score) AS Score1,

MAX( s2.Score) AS Score2

FROM Scores s1

JOIN Scores s2 ON s1.ranking = s2.ranking + 1

AND s1.Module_ID = s2.Module_ID

AND s1.Account = s2.Account

GROUP BY s2.Account, s2.EvalDate, s2.Module_ID

ORDER BY IdTime2 DESC

To clarify, I'm getting the scores (with the formula) before the JOIN. That way, you won't loose rows in your calculation. Using a self-join of the calculated recordset, you can have 2 scores compared at a time with infinite possible rows. The DENSE_RANK() function is to have a value to join with the previous evaluation.