Since we're not going to stop on this sales thing, I'd like to point out an error in the logic and part of why the numbers aren't measuring up.
If you're cancelling flagships that released prior to 2012 from the Android numbers and not including all of the other OEM devices and/or the 2012 devices for anything except the Note 2 and S4 then I'd knock off between 1/6 and 1/3 of the iPhone sales number. The reason for this is that Apple tells us total sales volume per quarter (in this case year) and not by model. The 2012 numbers include devices made in 2010 and 2011 and the 2013 numbers include devices made in 2011. In that vein, to make this an oranges to oranges comparison we'd either need to back those sales out (I was generous and assumed a 60% newest 30% legacy and 10% free phone distribution = 1/6 doesn't count) and we would need to include the Galaxy S, Galaxy Nexus, Galaxy S2 (and all variants), HTC Evo, HTC Thunderbolt, HTC Evo 3D, HTC Evo 4G LTE, HTC One X, HTC One X+, Sony Xperia lines from 2011 forward, all of Moto's devices (in US sales they didn't have a "junk" device, not sure about overseas but they're minimal), all of LG's devices plus all of the higher end models that exist predominately overseas from OEMs like Huawei, Xiaomi, Oppo, etc, etc.
Given that logic the best way to measure it would be to only measure new device sales by OS and then reduce Android's by the "junk" factor. Jeff's logic works in this example and could be applied to each quarter in the evaluation period (2012-now). We'd simply sum the total activations numbers for the 30 months or 10 quarters, apply the market share and multiply by the percentage that is "not junk".
If we give Apple Insider the benefit if the doubt, then 34% are "not junk" and therefore 34% of 81% = 27.54% and that is slightly over 2:1 (about 19/9) of Apple's ratio, for the quarter shown.
Either way, the list of devices that were sold in 2012 through 2014 is much longer than 10 devices. Those aren't even the top 10.