It was a "best practice" rule we used while I worked in the Lab at Anixter. It was just a generally accepted convention, but one that was pretty much accepted as "The Law."
When we did proof-of-concept testing for customers, or if we were doing anything related to performance of some networking component, staying within best practice convention (especially when there was no specific written rule) was very important.
I made a quick pass through the EIA/TIA 568 documentation and didn't see anything specific to minimum cable length. My other cable-related books & docs are buried and not easily accessible right now.
If this is a critical point for you, then I would take it within the context of the following:
The EIA/TIA provides specific thresholds related to transmission characteristics of Category-rated cabling. As long your configuration conforms to those specific characteristics, then the length of the cable (longer than 100M, shorter than conventionally accepted best practice)is moot.
The "100 meter" rule is merely a guideline, such that if you use Category-compliant components, and you stay within the recommended 100M overall length (5m stranded/jumper + 90m solid / "in-wall" or permanent + 5m stranded jumper), and all other rules (MBR, stretch, termination ...) are followed, your cabling system should perform at maximum performance levels.
There are always "I did it this way and it worked" (when "this way" is not necessarily by the spec)people: Just getting bits down the wire is not the goal.
For the time and money expended to create the network (at least in a commercial environment), you want to get the maximum possible performance. Operating outside the spec generally means sub-optimal performance (i.e., money wasted from being lazy, ignorant, cheap, or stupid).
You might be getting some bits down the wire, but not as many as you probably could get.
If I get time, I'll try to dig up the docs.
Good Luck
Scott
