There are a variaty of things I reutinely fudged with the CR/EL system.
1) NPC classed individuals: The general guideline, that character with X levels in a NPC class was CR X-1 seemed wrong from the offset. Even assuming that they were loaded with magical goodies, I couldn't see how they compared favorably to monsters of the same CR. I regularly treated the CR of an NPC classed individual as being only X/2. So those 6th level orc warriors were generally treated as CR 3, not CR 5. For commoners, I generally relied on X/4, so the 9th level commoner (really, these existed), was all other things being equal considered CR 2 (even though I never had to award experience for PC's killing a 9th level commoner, although it might have come up if I'd been able to run my adaptation of Isle of Dread).
2) PC classed individuals: Many of the PC classes were balanced on the assumption that they had great utility out of combat to match being slightly weak in it. Also, I rarely gave PC classed foes large amounts of treasure on the assumption that if I did, it would end up creating an arms race with the PC's who would acquire all that treasure for themselves. So that 6th level fighter was generally considered to be CR 5, not CR 6, and I generally rated the CR of mid-level NPC theives, monks, and bards at two lower than their level.
On the Ogre Mage argument, I'll chime in to say that I don't think any PC class can be considered in and of itself an associated class of the Ogre Mage. If the CR of a monster is primarily due to innate spell-like ability use and similar things that don't depend on HD, then anything that doesn't enhance its spell-like abilities doesn't directly enhance its effectiveness. In the Ogre Magi's case, that means IMO all PC classes are not associated classes. If I wanted to make a high CR Ogre Mage, I'd not do it by adding class levels. Instead, I'd directly increase its HD, spell resistance, regeneration rate, and innate spell ability - since these are the things that cause it to have its CR in the first place.
3) Large numbers of opponents without level independent attacks and 4 or more CR below party level: I considered a 'level independent attack' to be one that largely ignored the PC's defences so that its still at least somewhat threatening regardless of the level of the PCs. For example, most breath weapon attacks still do half damage to most characters even on a save, and ranged touch attacks ignore most PC's defences. If a monster didn't have a level independent attack, and its CR was ~4 or more below party level I found that it was a cake walk even in large numbers. As a result, I added only +1 to the EL when I doubled the number of foes, rather than +2. In this way, I could have the party face off against a fairly large number of low level foes without flooding the party with XP for an 'easy' fight.
4) Single opponents with low defences: I had very bad experiences with single foes in 3rd edition, and moved away from them the longer I DMed. Typically, if they weren't given some sort of overwhelming defence, they went down hard - often in the first full round of combat with the PC's and if not then in the second. That meant that they typically got off just one attack, which if it missed, turned them into an underwhelming cake walk. If you tried to use the normal monster design guidelines to solve this problem, you ended up with a monster with such overwhelming offense that it risked a TPK. I never found a solution to this I was fully happy with. Typically, I made sure that I'd found some way to up the defences in some manner and hoped for the best. I didn't adjust CR downward because focusing on a single PC could be very threatening if things went the monsters way.
5) Terrain: This wasn't really fudging, because the system mentions it, but I found that I needed to adjust things according to the tactical situation if it was really going to reflect the difficulty. I frequently awarded 200% or even 300% of the normal XP for having to fight in some horrible death trap I'd designed (in rapids, on top of a log jam, in a giant slippery funnel, in quicksand, etc.), or 110% or 120% for having to fight against well prepared foes (archers behind barriers), or just 50% or 75% normal for fighting a foe in terrain that favored the PC's (melee combatants at a distance across an open field).
6) Intelligence of the Foe: Similar to terrain, I tended to fudge the XP award by a small amount depending on how much of my cunning I felt justified bringing into play. Against say animals, the PC was facing a foe that would likely flee if things went the slightest wrong, and which didn't understand magic or missile fire all that well and which would probably be detered by encountering something even as mysterious (to it) as fire. Against zombies, oozes, or plants, the PC was facing a foe that would probably be no threat at all, if the PC could gain some separation from it and employ ranged attacks and which would continue to use straight forward ineffective tactics regardless of the outcome. Against a non-combatant foe (bullywog villagers), the PC was facing something that would spend more time running away than trying to threaten the PC's. So, I'd adjust XP down again if I thought the foe didn't represent as full of a threat as it might otherwise. A pack of wolves encountered randomly in the wilderness was quite a different thing to my mind than a pack of rabid wolves under the influence of some evil power or mind.
Even with that, I was often surprised by the outcome of encounters. The biggest variable was one that I couldn't really control, which was the actions of the players. Sometimes, they'd immediately seize upon a very good plan, and implement it immediately, and my monsters would go down hard. Other times, the party would become confused, party cohesion would begin to break down, individuals would panic, or do very foolish things and suddenly a fight I didn't think exceptionally hard turned into a full fledged rout. In this case, I generally didn't adjust XP, but I did tend to think harder about the factors that had lead to the parties confusion and sometimes factored them into my 'guesstimates' in the future.
![:]](http://www.enworld.org/forum/images/smilies/devious.png "Devious :]")
