That may be by design.
How do file cache measurements compare?
Actually, I'm quite sure it's by design. I took the Fusion apart because I'm constantly playing with stuff so I'll try to remember to look at that the next time I put together another.
Here's something interesting (quite interesting, IMH0) that I found:
Putting an SSD into an external USB 2.0 housing and connecting it to a USB 2.0 system, formatting it, and then booting it like that (non-Fusion) with Mavericks or Mountain Lion would typically take nearly a minute for everything to boot to the login screen. This is slower than if the drive was connected internally. The speed delay I assume is because the USB port is bottlenecking the SSD. If the same housing is using a fast Hitachi 2.5" drive (one of the new ones that uses AF format) the boot speeds are comparable. This is why I think I can confirm the bottleneck on the SSD is the USB port.
Now, if I combine the external USB 2.0 based SSD with an external (and slow) HDD to make a Fusion drive with it, the system would boot in just over 40 seconds, which is considerable faster. I rebooted like this several times to confirm it.
I have to assume that Core Storage/Fusion is implementing some type buffering or maybe using a much larger "virtual" sector on the Fusion drive that makes it behave faster. Obviously, that's a guess, but the speed increase is noticeable during boot. The point being Fusion seems to do more than just map data between an SSD and a hard disk drive. I have seen similar sorts of performance gains on striped RAID sets but I haven't compared them one-to-one.