sas/sata topology vs performance?

[Nicholas Leippe]

Regarding the 12-drive bay we have, I'm seeing some puzzling performance.
The drives are physically so in the bay:

e h k n
f i l o
g j m p

I configured mirrors across e+f and g+h, and both synced at the same, 
quick/expected speed. These are my four sas drives. The rest are sata.

If I configure a mirror with o+p, or m+n they both sync at between 50 and 
80MB/s--very quick and within expectations.

However, if I configure a mirror with drives i+j, it syncs at <25MB/s.
If I configure a mirror with either i+p or j+p, it syncs at between 50 and 
80MB/s--so it doesn't appear to be either drive i or drive j at fault, but 
their combination...

This doesn't make sense to me, especially where they are all point-to-point 
devices.

Is there some weird way that sas/sata addressing can affect this?