>>Obviously the 66 will still be 66, and transfering between the two will be limited by the 66.<<
Well no that is not the limiting factor here since the IBM nor the old Maxtor can neither push ATA 66 and certainly not ATA 100 to it's limits. The problem is the system actually has to pause and stop communicating with one device to communicate with the other, which has nothing to do with the transfer mode. Furthermore even on separate channels, the limiting performance of the Maxtor will be a bottleneck for data transfers from the maxtor to the IBM (depending on how the Maxtors read compares to the IBMs write speed). But in general for similar performing drives when the bus bottleneck is removed, this will always be the case cause the write capabilities will be slower than read capabilities. And this only applies to transfers between drives.. Any application run or data read from one drive it won't matter.
Hypothetically.. if the IBM or the Maxtor were able to push transfers beyond 66MB/s then that transfer mode would be a bottle neck and it would apply if you put them on separate channels, the same channel, different cards. It would always apply no matter what you did. But that's all hypothetical and the drive's can't, they don't, and the transfer mode isn't.
And these issues are beyond the simple question brought up, which is can a drive operate in ATA 100 mode on the same channel as drive operating in lower mode. The answer to that is emphatically yes. Even though this fact makes no real world difference in today's performance, in your hypothetical world where ATA 66 does create a bottleneck, it makes a very big difference as anything run just off the ATA100 drive in ATA100 mode would transfer beyond 66MB/s.
***Hey I run Intel... but let's get real***