They certainly shouldn't base their decisions on flawed information. The article you referenced doesn't provide any information on market share (i.e. the percentage of devices in active use), as it only presents information based on device sales, which has no correlation to market share. As an example, assume there are 3 people who purchase phones with different OS'es. Over a period of time, person A replaces his device 2 times, person B replaces his device once, and person C keeps his original device. According the sales estimates, OS A has 50% "market share" (3/6 total devices sold), OS B has 33% (2/6) and OS C has 17% (1/6), when the actual market share never varied from 33% each (1/3).
This information has been widely mis-represented in the online media by writers who do not perform their research, only reposting somebody elses's misrepresentation. The vast majority of market share articles today only reflect device sales, with another sizable chunk basing their market share estimates on web statistics, but this method also has flaws, as it assumes that all users use their smartphones in the same manner, and I have seen on many occations where my Windows phones have been mis-identified as iOS or Android by websites. I have been tracking the data from ComScore MobiLens for quite a while, and they base their market share estimates on carrier subscriber accounts information.