The Feedback module's 'multichoicerated' item type appears to use an incorrect formula to calculate average response values - this occurs when optional questions are included in feedback but not all users provide an answer.
When calculating the average for each question the module assumes it was answered by the total number of users that submitted a feedback response - regardless of whether they in fact they answered the individual question.
I've added a screen shot that should hopefully help communicate the problem here - it shows a feedback which 4 users have provide a response - question 2 has been answered by all 4 users (which shows a correct average value). Question 1 however, has only been answered 3 times and shows an incorrect average - it should be calculated as follows:
Question 1: (1x1) + (2x0) + (3x2) + (4x0) + (5x0) = 7 / 3 = 2.33 (correct)
... it appears to be calculated as follows:
Question 1: (1x1) + (2x0) + (3x2) + (4x0) + (5x0) = 7 / 4 = 1.75 (incorrect)
I've also attached a patch which should help fix this issue.