The .2 second improvements could be due to tweaking of the software. No customer would complain that Tesla made their car faster. Those .2 second differences don't necessarily point to inconsistency in testing methodologies or underrating.
The big .5 second differences between Motor Trend numbers and Tesla numbers obviously are a result of Motor Trend using 1 foot roll out and Tesla, at that time, and until they published a P85D number, not using 1-foot roll out.
Which makes my point.
The S85 is actually 0.6 second difference, which is far more than the 0.3-0.4 you would expect from rollout. The S60 if you take the 5.9 and the 5.1 measured on vbox (with rollout), that would be 0.8 second difference, again far more than rollout. So my point was that Tesla understating the 0-60 numbers of those models is something that is different from just rollout.
http://www.dragtimes.com/blog/tesla-model-s-60-kwh-vs-tesla-model-s-85-kwh-performance
As for software update, there was none for the S85 and P85 that Tesla said would boost acceleration (if it resulted in 0.2 second difference I'm pretty sure Tesla would advertise that; they certainly did for the 0.1 second on the P85D). They just simply tweaked the advertised number. The measured numbers did not have any notable changes.
For Tesla to suddenly switch to using 1 foot roll out, without and indication whatsoever that they had changed their testing methodology, was misleading. Any reasonable person looking at the Tesla-published P85 0-60 time of 4.4 (taken from your post--I haven't looked this up) and the P85D 0-60 time of 3.2 (the original time Tesla published) would think the P85D was 1.2 seconds faster 0-60. They would have had no reason to believe Tesla had suddenly started using 1-foot roll out. And they would not have been able to compare magazine reviewed times for the two models, because the P85D wasn't available yet.
I have no issue with Tesla deciding to use 1-foot roll out, since a lot of publications do, and many manufacturers do. My issue is with the fact that they did not give any indication whatsoever that this was what they were doing. I feel strongly that they had a responsibility to do that, especially since this was a CHANGE from what they had been doing previously.
MotorTrend had their first test in 11/3/2014 of the P85D where they got 3.1 seconds and they were very explicit about the rollout (they typically don't even mention it):
"the Tesla punches the clock at
3.1 seconds, a tenth quicker than the Audi (as well as the McLaren F1's accepted time -- all of these after subtracting the customary
1-foot rollout)."
http://www.motortrend.com/roadtests/alternative/1411_2015_tesla_model_s_p85d_first_test/
12/5/2014 was when the first delivery happened, so the tested numbers were all there already for comparison before the first delivery occurred!
http://insideevs.com/worlds-first-tesla-model-s-p85d-owner-gets-interviewed/
Comparing Motor Trend numbers including rollout: S85 to P85 had a
1 second gap (5.0 vs 4.0), P85 to P85D had a
0.9 second gap (4.0 vs 3.1).
Comparing Tesla advertised numbers during P85D launch: S85 to P85 had a
1.2 second gap (5.4 vs 4.2), P85 to P85D had a
1 second gap (4.2 vs 3.2).
So there was only a 0.1 second difference between the advertised and measured gap between P85 and P85D and 0.2 second difference for S85 and P85 (actually more difference, which would seem to disprove the theory of those models tested consistently without rollout). Again, it is not a 0.3-0.4 second difference as you would expect if rollout was the main thing at play here.