While I do agree somewhat I think if they really cared about that they would have ported it back when Windows 7 had >50% marketshare not too long ago rather than now when it's been in free fall for a while, has around half of the users of Windows 10 on Steam and now only really still has a stronghold in markets like China with even these markets deep into the sharper transition western markets saw over the last 18 months. Personally I don't think the existence of DX11 in modern games is entirely down to Windows 7, many people still have Kepler based hardware that doesn't play nicely with DX12 at all while even Maxwell generation hardware is particularly well optimised for the HLSL style APIs and Pascal only got as far as not having a massive penalty from many low-level API features compared to the well optimised HLSL driver stack, with few benefits from the low-level approach on any NVidia hardware until Turing. While we're now starting to see good implementations of DX12 that improve CPU use and such, creating a good DX12 implementation in a game is orders of magnitudes more work that for a comparable DX11 one, and it made sense developers had to take time to optimise those implementations well especially when most PC gaming hardware didn't really suit it yet.
Around 70% still use NVidia hardware that will gain little to no benefit from a low level API, or around 3 times more than those limited from using DX12 as a result of their OS, personally I think the the significant improvement in DX12 libraries and work done on them recently has been a major contributing factor to its viability as this is what will allow many developers to practically use the API, and if it weren't for the consoles there's a good chance we'd have never really seen DX12 go anywhere at all as NVidia could have dug their teeth in with their own quite contrasting design philosophy.
