The question keeps coming up whether the QPDs were centered during these calibration runs, and whether the amplitude of the excitation was too large, pushing the spots outside of "centered" during the calibration excitation run.

Here, I attach a trend of the SPI optical levers' QPDA and QPDB during Jim's excitation. For a little more detail, the excitation points used were at the input to the isolation filter banks;
    H1:ISI-HAM2_ISO_RY_EXC_DQ >> PIT excitation
    H1:ISI-HAM3_ISO_RY_EXC_DQ
    H1:ISI-HAM2_ISO_RZ_EXC_DQ >> YAW excitation 
    H1:ISI-HAM3_ISO_RZ_EXC_DQ
and thus the units of the amplitude of the excitation he quotes are in nanoradians.

The response channels that Jim used are 
    H1:SPI-H23_OL_QPD_A_PIT_OUT_DQ
    H1:SPI-H23_OL_QPD_B_PIT_OUT_DQ
    H1:SPI-H23_OL_QPD_A_PIT_OUT_DQ
    H1:SPI-H23_OL_QPD_B_PIT_OUT_DQ
As discussed in, e.g. LHO:91210, these channels are records of the normalized spot displacement at the face of the QPD. Thus, the units of these channels are [V/V] ***, because we've chose to calibrate each QPD segment into ADC input voltage (in [V]). These are the units in the denominator of the transfer function Jim quotes "?" above.

The QPDs were relatively centered during these measurements (mean each less than +/- 0.02 [V/V]), and during the excitation, the response amplitude did not exceed +/- 0.15 [V/V].
As such, we can safely use the linearized version of the error function -- erf(z) -- to interpret the calibration of the optical lever QPD as linear (again, as discussed in, e.g. LHO:91210; erf(z) ~ 2 z / sqrt(pi) for  z less than ~0.5 ).

*** Because it's the ratio of identical units, the normalized signal will sometimes be labeled with "dimensionless" units, or "[radians]" I've even seen "[abs]," which may be a version (or typo) of "[arb.]" for "arbitrary units" which is a misnomer for "this signal is the ratio of two identical dimension-full signals, so *technically* you can label it with any ratio of identical units; [V/V], [m/m], or [elephants/elephants]. Don't be fooled. This signal is the ratio of dimension-full signals which started out as having units, and while arbitrary, we *chose* those units to be ADC input [V].