Well, that statement can perhaps be true for new and innovative features.
But a lot of the issues are patterns that have been established for 10-20 years in the GUIs of Windows, GNOME 2, and others, and get dropped with no consideration of the obvious fact that everybody expects them and has muscle memory for them, and that they are often very good ideas.
Things like Ctrl-C/V/X doing copy/paste/cut (since OS/2 I think?), Alt-Tab switching between windows and not between application (since at least Windows 3.1 in 1992), non-global menu bars (since before Windows 3.1 in 1992), having a taskbar on the bottom (since Windows 95 in 1995), launching apps by clicking on bottom left or pressing the Win button (also Windows 95), a clock on bottom right (also Windows 95), the ability to search pressing the Windows key and typing (since Windows Vista in 2006), a tree view in the file manager (since before Windows 3.1 in 1992), a bookmarks bar in the browser (since at least Internet Explorer 4 in 1997) etc.
These are things that any desktop environment that targets people who were previously using Windows MUST have as an option, or you'll make an horrible impression right away.
Furthermore, since GNOME 3 was supposed to be an upgrade, breaking the assumptions that GNOME 2 users made was also utter folly (and you broke Alt-Tab, broke the window list, the clock, the menu, etc.).
Note that most people never experienced an abrupt change in desktop environment on the same device (since they always used Windows, which barely changed in the fundamental UI paradigm since Windows 95), so forcing them to accept them when switching to an obscure platform like Gnome with less than 1% market share is folly, they'll just say "WTF is this shit" and go back to Windows; there's no way most people are going to bother to learn GNOME 3 if takes a lot of effort without spending millions that GNOME doesn't have in creating marketing hype, or tying it to hardware.
So, it's for EXISTING features that you must ask users (and actually there's no need to ask, just don't mess with them...), but you can definitely freely invent new optional stuff, and in fact if you manage some "killer improvement" that's fantastic.
Also, it's obviously OK to enhance existing features, as long as when the users performs some action, he gets a result which is recognizably similar to what the old version did (e.g. he presses Alt-Tab and gets some sort of display of all windows, and pressing Tab again rotates between them).
As an anecdote, personally I tried OSX as a Windows and later Linux user and found it horrible and unusable precisely because it breaks pretty much all of those conventions (including Ctrl+C for copy, they use Command+C instead, seriously!), and doesn't even have options to fix them; it was so terrible that I deleted my OSX VMware virtual machine in disgust rather than keeping it around in case I needed it.
That's an impression you don't want to make.
Conversely, a Windows user starting to use GNOME 2 is usually greeted with an horribly wasteful configuration using 2 horizontal panels (God only knows why distros usually shipped that as the default), but there at least it's very easy to remove the top panel and fix the bottom panel to be taller, and have a Gnome menu on left, window list in middle, and clock on right (although you need dockbarx or similar to emulate the Windows 7 taskbar).