I always recommend a 20 starting stat, and always raise it and your secondary. Of course, most people end up watering that down and pick a 19 or even an 18. And the character optimizers find lots of (valid) reasons to deviate.
And that's kind of the point.
People who are interested in the game mechanics can figure out for themselves if losing attack/damage is worth it. They'll ignore the advice and there's no problem. On the other hand, people for whom character optimization is just a required filler for the game, people, like this:
Boy you should watch your players' chars more closely. The problem is not that he is playing a Baladin (STR/CHA) but that he doesn't know how crippling those ability scores are. (Even worse he might be using implement and weapon attacks and maybe the weapon and/or the implement is outdated pluswise.) You really should explain the "problem" to your player and help him rebuild his char.
Actually, I looked at the paladin on the CB, and she hasn't bumped up str and cha every time she gets a stat bump. However, neither the weapon nor the implement (she doesn't have a lot of implement powers anyways) are outdated. Those are things I can handle. It's weird, I've never been shocked that someone wasn't enough of a munchkin.
Well, if they just pick a 20 and raise it, they'll miss a few powerful feats and lose a (usually minor) bit of secondary power, but that's
OK. They will, however, remain playable and reasonable relevant in comparison to their friends that do optimize.
Recommend a 20/(14 or 16) starting stat and the primary/secondary stats should
always be raised. Not because it's best, but because it's safe for those that don't care - those that care can figure it out themselves.
What I did in subsquent encounters was lower the monster's defenses as a temporary stop gap for everyone. I may continue this rather than try to play someone's PC for them. I mean, their just trying to play well-rounded characters. That seems like something I should encourage.
A big part of 4e is tactical combat. Munchkin is a pejorative term; I think you
should encourage everyone to make an effort to participate - and that means a bare minimum of optimization. Think of it this way: doing otherwise undermines the balance and gameplay of this major component, and that's not fair to those that
are trying to play. To me it undermines flavor too; I expect bad guys to actually want to win and if your PC is having great character development in several directions but just isn't developing into an adventurer, the PC should retire or not be surprised when he dies. But then, I like slightly harsh campaigns

.
Unless all the players have similarly gimped attacks, you should really try to fix the problem since it's just going to diverge and get even worse. 4e doesn't work well if people distribute stats based on some abstract notion rather than as dictated by the build.
A few things you can do:
- The player may just not care about mechanics and may be happy to let you fix them.
- Or, if he does care, it's OK to put a little pressure on him (i.e. have him less effective in combat) - it's a cooperative game and it's his job too (not just yours) to be a team player. By insisting on a build that just doesn't conform to 4e expectations he is putting an extra burden on you and the others.
- You could also make a special exception for his build and just raise his stats arbitrarily: the CB will happily accept (house-ruled) stats that are even over 20 at level 1, so whatever you think is reasonable goes (just make sure it's still less powerful than those that are trying to play by the rules, to be fair - at least 1, maybe 2 lower primary stat modifier should suffice to avoid encouraging others to be similarly difficult). Call it a 4e class design error; it doesn't matter.
- I also advocate the house-rule whereby all stats rise at levels 4/8/14/18/24/28 and not just two. It avoids these tricky player errors, it avoids the socially tricky moment of you implying "your character sucks", it avoids skills and NADs from diverging, and it's easy to implement in the CB (just don't use the ability score increase boxes but instead raise all scores by one). Finally, it fixes a few of those MAD builds that ask a player to shoot themselves in the foot. A character with low primary stats will still be weaker, but at least he won't fall ever further behind. Also, if someone doesn't preplan his character for 20 levels, he'll still eventually get the stat prereqs for most feats, which is nice too. Note that you may want to tweak the barbarian's agility since that feature relies on light armor wearing barbarians that don't get Dex rises.