Even after 3.e3 White is not forced to accept a transposition to the QGA: he can continue to hold back his d-pawn and develop instead with b3, Bb2 etc.

Remember first of all that 2.c4 isn't even a gambit.  White could play 3.Qa4+?! and then Qxc4.  It's not the best way to play though.

My shortest win ever went like this:  1.Nf3 d5 2.c4 dc4?! 3.e3 Nf6 4.Bc4 Bg4? 5.Ne5!? Bd1?? 6.Bf7#.

I don't think you really need to worry too much about 2...dc4 because it breaks several opening principles at once:  Don't waste time grabbing unnecessarily; develop toward the center with your pieces; don't open up the game unless you have an edge in development.  2...dc4 isn't as bad as I make it out to be, but it should be one of the reasons for White to play the Reti!

The pseudo-gambit 2.c4 invites Black to make a mistake.  This is part of why such strong GMs as Kramnik play this move order to get into Queen's Gambit lines.

Certainly seems like the right kind of plan.

A slight warning though: iirc from analysing this with a computer a while ago there are some rather annoying black ideas where he plays Nf6 hitting e4 then follows up after e5 with Ng4 ^ either Nxe5 & Qd4 (forking N & R) or simply Qb6 attacking f2.
It is all quite artificial and I suspect white can cope somehow but it needs working out.

Another very important line to look at which I believe I found in some old edition of MCO is 3 .. e5, which in some ways seems more logical than c5. Since Nxe5 Qd4 (^ Qxe4+) looks like nothing for white, you probably have to try 4 Bxc4 Nc6(say) etc. 
I guess white's ideal plan here would be to make d4 work without playing d3 first, as a pawn sacrifice if need be. If white has to slowly prepare d4 I can't see black being too worried.
(Still plenty of play in the position of course.).

Sorry if this is all a bit vague - I last looked at this a while before the book came out. And was pleasantly surprised to see I wasn't totally crazy when it did