Tuition begins to matter when students lose funding. Then it's no longer "funny money." According to the Ph.D. completion project, between 50–60% of students that enter a Ph.D. program finish it by the 10 year mark: http://www.phdcompletion.org/quantitative/book1_quant.asp
Most programs guarantee funding only through the first 4–6 years. Some programs make students compete against one another for quasi-guaranteed funding after a certain point.
A substantial minority of graduate students are either going to quit the program with only a masters degree to show for it or are going to end up paying tuition. Usually they can reduce the amount of tuition paid by enrolling as a "non-resident student" or something similar. But the tuition rate (and how long they will cover tuition) isn't something that can be brushed aside, either by students managing their finances, or by departments assessing their budgets.
Also, are you saying some PhD's take 10 years to complete, not including undergrad/masters? My understanding was a PhD on its own was typically a 3-6 year endeavour. I know it's a difficult project to manage but 10 years seems long.
When I was applying for grad school I got a letter offering admission from some school that basically said: "we'll pay you $20k in stipend, plus tuition is $40k, so really this offer is worth $60k".
I think they may have added another couple thousand for health insurance as well.
(said a fellow grad-student)
In my field, no competent grad student paid tuition, and it wasn't a meaningful measure of value or cost. Nobody would have accepted an offer from a grad program that didn't provide a full ride. So tuition was basically funny money, like a "list price" that is always discounted.
My research work brought in enough grant money to pay for my stipend plus equipment and other expenses. If anybody was taking a loss, it was the National Science Foundation.