Probably a lot of that extra money went on bureaucracy, meetings endless paperwork and mind-changing (leading to spec changing).
However, according to the ESA, "The total development costs for the ATV amount to approximately 1.35 billion euro. This includes the prototype (ATV 1 Jules Verne at around 1 billion euros)".
I don't think that the development costs of Dragon nor the launcher have been released. Nasa paid less that $1bn, certainly, but they certainly weren't the only contributors, and don't own the rocket or Dragon (they are just buying them like you buy a TV, and what you pay for a TV doesn't cover the design costs either). NASA is paying at least another 1.6bn to launch 12 missions, and with the re-usability that SpaceX is looking at, their costs here will be lower, and some of the profit will probably pay off some more of the initial investment.
As for the ATV, it already has life support and is supposed to be fully human rated (although not launched with humans on board). The ATV guidance system is also a fully autonomous docking system, whereas the Dragon is told to fly in a particular way and is plucked out of space using the ISS arm. The ATV pressure vessel is also a bit bigger than Dragon's, holding 7 1/2 tons of pressurised cargo vs. 3. (also 4T of ATV fuel can be used for reboosting the ISS), and the ISS stays docked for 6 months giving a sizable increase of habitable space, compared to the 2 weeks (and smaller space) for Dragon.
The proposed ARV variant that ESA has been umming and arring over for far too long would also solve the re-entry problem (but losing about 1/2 the pressurised volume/cargo capability in the process). However Europe needs to do a lot of work to get the knowledge / experience to do that, not having done that already.
All that said, for it's job I like the Dragon a lot. I like the way it doesn't need a huge fairing over it at launch; I like the way it is designed for hatches, windows, seats and integrated launch abort system from the outset. If it can be caught and manually docked, that simplifies the AOCS a lot, and seems to be acceptable (although how it will dock with a bigalow is another matter).