Can't the SDRAM bé clocked at 133MHz to match the 1:2 ratio with the CPU like with 60Mhz.
We did indeed test the memory and the drivers at 133Mhz successfully during the validation of the hardware/board design. The challenge is making the CPU happy with the returned data and and making sure it arrives on time. Now the faster you run the clocks the trickier this gets. At those suggested frequencies you get into margins of just a few nanoseconds for things to arrive on time. If you are impressed by zeros a nanoS is 0.000000001 seconds. And that must be correct for all 4 memory banks simultaneously (to give you this amazing 254 MiByte fastram that the card offers). With a CPU that thinks 15 nSec is already fast according to its datasheet (to get that into Mhz do 1000/15 is only 66 not this magic 133 Mhz so much desired, so yeah.. there you go technical details 
Now our top priority is stability of the product first, insane speed comes second (or maybe third and the other features second.. dunno something like that 
So we are currently working on the IDE speeder (making writes faster), the software tooling and bugfixes.
Only then we can look at some improvements in memory performance and this might indeed be one of areas we can try to improve on. We know performance is important too, but preferably without causing instability as that is a bad trade-off.