NASAR could build cars cheaper than they race on Sundays too. That doesn't mean they would get the job done...
The primary reason most clone workstations do not make good servers is simply the components they use. They are designed for light to moderate use by a single user. They rarely include the kind of robust performing parts that come in servers - redundant power supplies, RAID (or at least SCSI) drives, industrial tape drives, and of course very carefully chosen video, nic, etc parts to make ABSOLUTE certain they work reliably, quickly and above all 100% compatibly with the server OS.
You *CAN* build this kind of machine...but you'll find that it suddenly doesn't cost what a clone workstation costs...it costs close to what a server costs. Worse still, as a clone box, you'll run into nightmares when it comes to support.
This is the most critical issue of all...because when there is a problem with a workstation, it affects one user's productivity. A problem on a server can affect the whole company. God help you the day the mail server fails if you don't have your disaster recovery and support ducks in a row.
I know this isn't what you asked for: references. This is just the voice of a long time admin stating what most of us consider common sense. You get what you pay for 99% of the time. Your production servers are critical to your company. Some times what saves you a few $$$ to start with can wreck the corporate bank account, your job (and reputation)...and perhaps the company in the long run. Lose a few years of records or critical accounting files during an audit...and you'll be asking "would you like fries with that" about the same time your boss will be facing an inquiry. How many dollars is avoiding that sort of disaster worth to your company?