Well, okay. I built a quick little monte-carlo testing program in VB to try out some variable adjustments.
I built a level 10 party along the same lines as Stalker0's example party -- one awesome skill guy, two moderately-good ones, one moderate-low (-2 below the moderate-good ones, equivalent to an armor penalty or lower ability score), and one awful (untrained and using an non-focus skill). The 'awful' guy always provided assistance (I really can't use the term "guiding light"...) to the second-lowest guy, so his checks don't count towards success or failure but might give the moderate-low guy a +2 to bring him up to moderately-good.
The program takes into account one Bold Recovery attempt per turn, made by the guy with the highest bonus. Technically you could have multiple bold recovery attempts made by different players if you had multiple final failures in a single turn, but that would add a layer of complexity -- so just assume they always fail those attempts. Also, this whole thing ignores the use of Heroic Surges, so the actual success rates will be slightly higher if the players want to spend resources, and it ignores any skill-boosting utility powers, which could theoretically convert a "moderate" to a second "good" skill user, or boost the low guy to a moderate. Whatever. Those are too situational to try to figure in.
In any case, this is what my little die roller came up with:
Given an on-level challenge the success rates were as follows:
Comp 1: 75%
Comp 2: 66%
Comp 3: 60%
Comp 4: 55%
Comp 5: 50%
The last two don't seem too awful; a second good skill user or lots of surge use could easily kick those up quite a bit.
Adding 1 to the DC (or, equivalently, running into a challenge designed for a party 1 or 2 levels higher) has the following success rates:
Comp 1: 66%
Comp 2: 53%
Comp 3: 45%
Comp 4: 39%
Comp 5: 33%
So I definitely wouldn't recommend using the higher complexities against a lower-level party. Adding 2 to the DCs (or using a party 3 to 4 levels lower) results in a massive drop in success rates, down to around 50% for complexity 1, so I really wouldn't do that.
Subtracting 1 from the DC (equivalent to hitting the challenge with a slightly higher-level party) provides these numbers:
Comp 1: 83%
Comp 2: 77%
Comp 3: 74%
Comp 4: 71%
Comp 5: 68%
So that's just fine even without any surges.
Given those numbers, I think it would be appropriate to calculate XP as X+1 monsters of that level, where X is the complexity -- thus complexity 1 is the same as 2 monsters, complexity 4 is equivalent to an on-level Solo (requiring significant resources and maybe some luck to beat), and so on.