I have been benchmarking all the queries on our site and making improvements where possible but this has got me stumped - I feel the changes I've made to a particular query should have a significant impact on the speed.

The original query was 'using where; using temporary; using filesort;'
The new query was just 'using where;'

Initial tests last week indicated a significant improvement but this week this has not been the case - I rewrote the test to alternate between the 2 queries in blocks of 10 to minimise variation (query caching is off ofcourse) but statistically speaking, even at a 95% confidence interval, theres no significant difference between the queries.
Now I feel there should be. Anyone had a similar issue?