Ok I played around with my 85 for like 15 minutes going between spot and point area expansion and both of them locked on pretty quickly, definitely not a couple seconds. The 85 isn't a very fast focusing lens of course, but it's definitely faster on the 5DIII than the II. The only time the focus was very slow was in live view mode, but through the viewfinder it was very fast. If this was your experience with the 5DIII you may want to play around some more, even after 2 months I'm still amazed by how often the camera just nails the focus. Especially in Servo mode with fast moving subjects, but one shot is a walk in the park. In my mind comparing the AF between the II and III is like a joke, I feel like the 5DII might as well not even have AF after using the III.
I wouldn't be too discouraged, it may just be something minor.
I wouldn't say the 5D2 AF is necessarily a joke when talking one shot center point. Indoors in a bright kitchen at night 5D3 did have a quite better hit rate than 5D2 with a 24 1.4 at 1.4 in two tests. But in a super dark room at night they basically had the same hit rate, 5D3 not better at all, both missed about 25% of the time by a fair degree, the absolutely 100% dead on rate was some deal lower than that, very tricky subjects and conditions though.
Overall the 5D3 is solid for one shot, sometimes noticeably better other times about the same rarely a trace worse than 5D2.
The 7D is the one that gets dusted by both for super low DOF fast aperture one-shot accuracy. Both my 5D2 and 5D3 easily beat it. 5D3 has best one shot out of 20D,40D,50D,7D for sure and better than 5D2 also, although only in some circumstances is the difference to any noticeably degree.
The 7D and 5D3 can drive AI Servo on 300 2.8 quite a bit faster than the 5D2.
I actually find the 7D, for soccer, so far, doing worse than the 5D2, once by quite a lot. It seems like the 5D3 will be the best, although the 5D2 can actually track quite well at times so long as the players are not really close and moving in certain ways (I also find my 5D2 does soccer radically better when the hidden AF assists are turned on, if you leave them off, then it suddenly does no better than an xxD at best in my experience; when players are not too close the 5D2 can almost do soccer as well as 1D2, when they are very close and moving fast then it seemed to get beaten by the 1D2 for sure) and I did notice the 5D3 do a few super weird things, like tracking a goalie slowly jogging out to toss out the ball with AF points kept totally on the goalie the whole time and a few times for one frame it suddenly grabbed the background stands at least 40' behind before jumping back perfectly on track the next frame, there was absolutely zero reason for it to have done that, very, very odd. It also once slightly missed initial focus but then instead of correct properly it someone jumped to like 5' in front of me to grab a railing only fractionally in the bottom of the frame and not even near any AF point. Sometimes it would slightly lag behind or jump ahead of a player too, although mostly the frames were still usable if not really quite what they should be but then again with the other cameras those misses were usually enough to make the frame unusable. When it did suddenly jump way out of wack the next frame was always perfect, with the others sometimes the next frame would be a touch off still.
I really need more time with all of them for soccer though to really say, lighting can make a huge difference and the shooting circumstances and lenses were not always the same so far. And one game is hardly time to fiddle with all the settings. I basically just left it on default soccer mode.
7D seemed pretty decent for surfing though.
Sometimes the 5D3 doesn't really seem to AF any better than the 5D2 although other times it does (talking about keeping AF to center point, certainly if you go away from center point then it's much better 90% of the time). It seems to AF better than the 7D and xxD. So I'm pretty sure it will be the best of all the non-1 series, with the 5D2 for many situations the next best. How much better overall than the 5D2 I can't quite say yet. Sometimes it has seemed to do quite noticeably better, other times just barely matching the 5D2 (but given that I have yet to hit any circumstance with the 5D2 seeming to ever do noticeably better than the 5D3 and I have sometimes hit upon the reverse, it seems clear the 5D3 is better, hard to say how much yet though).
Interestingly, it was in the most brutal lighting and subjects of all where the 5D2 seemed to give the 5D3 a run for it's money with 24mm 1.4 shooting. The 7D did way worse in that situation than either.
In low because it's indoors but a well lit room nonetheless the 5D3 seemed to pull a fair amount ahead of the 5D2 with 24mm 1.4 shooting. In this case 5D3 was kind and then 5D2 next but noticeably behind and then behind the 5D2 were 7D and xxD at the back of the pack. So with somewhat better but still night indoor lighting the 5D3 pulled ahead but not at all when the lighting was truly abysmal indoor type.
Both those tests were center point only.
5D2 outer points fail to lock on many subjects which can be annoying, 5D3 ones tend to lock. Having compared accuracy when both lock yet though. My feeling was the 5D2 and 7D had a bit more accuracy when the outers locked than the 20D,30D,40D,50D although the 40D,50,7D would lock much more easily, especially the 7D, than the 5D2 outer points.
Need to use it a fair amount more to really say.