I run taking into account all 1167 observations (I do not try to identify outliers):
- 1164 obs from 1978 to now
- 3 obs taken in 1956
.
RMS error, with all perturbers on: 0".589
residuals for the 1956 obs, around 1".8 in RA and dT around 4 minutes

RMS error, as before but without Venus as a perturber: 0".598
residuals for the 1956 obs, around 0."2 in RA and dT around 30 seconds

So the counterintuitive effect is not found for the whole RMS error
but just on the residuals of the 1956 observations.

As you say, this effect is probably found because I do not throw away some outliers. Furthermore, paying too much attention on the residuals of just a few obserbations may be a wrong approach!

Should this prove to be one of those _rare_ situations where the effect is real... as you say one can be sure that Venus was there in 1956! :-)
...
so I am just wondering whether there is some physical explanation: for example, the apparent worsening effect that Venus has on the 1956 residuals is due to the fact that we do not take into account other perturbers (different from CPV) that counterbalance this effect.

Cheers,
Alessandro