The junk filter is a 'learning' filter. It gradually learns via what you decide to become junk. After a while it learns certain qualities of the same type of junk file are the same & thus designates such files as junk when your search results are returned.

I personally recommend having lowish set strength of Junk filter. Also, choose to set to show junk at bottom of search results. In that way you can double-check to make sure no files have been mistakenly designated as junk by the default Junk filter settings or settings you have created along the way of training it. It's not a 100% perfect tool, and should always be treated with some care & thought when used.

Here's an example of junk file sizes & for that matter words also that are typically junk files & should be designated as such: Virus thread: Typical virus spam file sizes - Make a note of them (click on blue link to see)