24 pointing data with RxA/ACSIS were collected in rather poor weather (0.15 < tau < 0.25), during a dedicated pointing run lasting about 3.5 hours. Removing collimation offsets of (CA,IE) = (-0.4",-0.0") -- and one clearly deviant observation (so N=23) -- gives the residuals shown below:

Raw rms residuals in (dS,dZ) are (1.8",1.7"); a remarkable confirmation of the utility of the new model despite the fairly marginal conditions.

Apparent remaining systematic trends in the residuals (eg dS -vs- Z - centre-left) are slight but still entice an optimization of the 7-parameter model:

       coeff       change     value    sigma

  1     IA         -3.595   -252.45    6.544
  2     IE         -2.097    -26.53    1.165
  3     NPAE       -7.456    +53.73    8.412
  4     CA         +6.894   -189.64    9.907
  5     AN         -0.438     +8.45    0.608
  6     AW         +0.481    -14.86    0.517
  7     TF         +0.134     -9.57    1.761

-- a 'tweak' apparently, given the low significance of any of the changes suggested. However, the new model should yield the following distribution of residuals:

- and rms's in (dS,dZ) of (1.5",1.6"), which are not so insignificant.
The new model was installed at 10:40 HST, 28 Feb 2007.

[Later: It should be noted that a concatenation of the 30 data from UT 20070221 with these (23) new data confirm the performance of the model (with (dS,dZ) = (1.9",2.2")) without justifying the update described latterly above. A full 7-parameter optimization does not further improve the residuals' (dS,dZ). However, we will proceed to use the very latest model, and hope to see (1.5",1.6"). If we see only (2.0",2.0") then the model is still OK -- and presumably as good as that produced on UT 20070221.]