If you have the following configuration in a job :
Timeout as a percentage of recent non-failing builds 200%
Number of Builds 5 (Number of builds to consider computing average duration)
Timeout minutes 45
Then, you might have one successful build for a job, but doing nothing (no build necessary), thus with very little execution time (less than 2 min) .
Then, you will get timeout errors when the time elapsed was 3 min and thus you will not be able to run successfully the job when the build was necessary. You needed to remove the successful execution.
A better behavior would be to use the average execution time only if number of builds is already bigger than the defined number of builds (5 in this case).
Before you have reached that number, the plugin should use the default defined timeout (here 45 min).