In the optical lever scheme, a beam is sent to the center of the the mirror with an angle α.

A rotation of the mirror around its vertical axis (a yaw) of an angle β, corresponds to an angular displacement of the reflected beam of 2β.

This produces a shift of the beam's position on a PSD placed at a distance L from the mirror of

X = L tan (2β) ≈ 2Lβ

Now, if the beam impinges (with an angle α)  at a distance d from the center, the shift on the PSD becomes 

X^d =  L tan (2β) + d (sinβ/cos(α+β)) ≈ 2Lβ + d β (1/(cosα - βsinα))

In the case of the beamsplitter, the optical lever beam is currently sent with a very small incidence angle (α ≈ 0)  and impinges at a distance d (in the orizontal direction) of about 3.5 cm from the mirror center.  The distance between the mirror and the PSD (arm of the optical lever) is about 60 cm.

The relative errore due to the mis-centering is   (X^d-X) / X = d/(2L) = 3%

which is small. So it seems not so crucial to hit the center of the mirror. 

For what concers the pitch, I would say that as long as the beam is on the orizontal axis (as in our case), its mesurement is not affected by the mis-centering.