Moreover, for all of the posters on this message board who have been down-right effusive about the "against top teams in the MAC stats" you would think at least one could at least try and come up with a quantifiable justification for excluding a majority of the data from the very data set they use to "prove their point".
Ok, I'll take the bait. And I'll do so in the style of the OP...
Lets imagine that you work at Apple back in the mid 2000's and you are tasked with testing the accuracy of the GPS receiver in the iPhone 3G prototype. You're given the phone (don't lose it!), a secondary $10,000 dual-frequency GPS receiver that can provide you the "ground truth" (don't lose it!) and some software that records and displays the difference between the phone's position and the ground truth (or the "error").
You're driving around the suburbs in Cupertino and so far so good..the phone is providing the expected accuracy (10-20m or so??). You drive through Mountain View, Palo Alto...you stop in Palo Alto to get some coffee at Philz (a Mint Mohito) and suddenly the phone rings. Its Jobs himself! He needs you to drive into San Francisco and test the GPS there because he frequently gets lost navigating the city. You jump back in the car and head off.
You finally get to San Francisco (stupid traffic!) and you enter the Financial District...and suddenly the software is registering large deviations from the ground truth: 150m, 345m, 501m, 237m! You start to panic. Did I break something? Did I spill some Mint Mohito on the prototype? Oh lord Steve will have my head for this!
You drive back to Cupertino and tell the engineering staff and Jobs you broke the phone and submit your letter of resignation. They laugh loudly and inform you that nothing is broken...you just encountered the classic Urban Canyon phenomenon where GPS performs poorly when the receiver is surrounded by large buildings. But Jobs has a look of consternation across his face...
***And here's the important part.****
Jobs is concerned because even though only 27% of the GPS data showed significant deviation from the ground truth, that 27% demonstrated a significant flaw in the receiver. It's fine that 73% of the data showed the receiver performing well, but that was in a not-so challenging environment. He needs the GPS to work in places like downtown Chicago, Manhattan, Shanghai, etc... So he sets his engineering staff to work on reducing the error.