One other theory is that the chip that features in the original PS3 was a stop-gap, thrown in oversized processor with most of it's functions ignored, and the new chip isn't a 65nm varient but a cut-down processor that's lost all the unused transistors from the original chip. In such a case the ratio of shrinkage would just be coincidental. That begs the question why such a stop-gap chip was used though.
Cost and time.
Developing any chip on a high end process is very expensive ($15million+) and takes a long time to do. If Toshiba were developing an I/O chip anyway (it looks perfect for a TV) then Sony could use it and put their efforts into developing a cost optimised 65nm part. They may be taking a financial hit by using it but the 90nm PS3 was probably never meant to be produced in large numbers anyway i.e. 6 million isn't very big compared to the 100M the PS2 sold...
Many embedded processors have all manner of weird and wonderful I/O ports and it's highly doubtful that anyone uses all of them. It's cheaper to develop a single one size fits all part than a series of specific chips.