Status: I’m fairly confident of validity and completeness
Over on G+, Dan Maruschak laments that many rpg groups will not or cannot follow rpg rules as written. I said
“In most life contexts, believing written rules and procedures is of dubious value. When you enter a new community, you’re usually better to follow what others do rather than what the rules say you should do. The real rules — those that most people follow in practice — are unlikely to be exactly those that are written down.”
to which he replied
“Which is why I find it so aggravating that that “follow the written rules of a game” is such a hard sell. People can get no-rules everywhere else, why demand no-rules here, too?”
In other words, Dan is asking why people don’t take rpg play as a chance to enter a well-defined magic circle — to play, for a change, in an environment defined by rules.
I think the following reasons are all factors here:
Because they can
First, players ignore rules because they can, without the activity breaking down completely. As David Berg puts it, over on Story Games:
“… Going on adventures in your imagination is inherently fun, even without rules, in a way that pushing cards around a table is not. If you sit down to play a card game, you’re a captive of its rules for as long as you’re playing, so you’d better either try your best to get the most out of them, or decide conclusively that they’re not for you. In an RPG, there’s no such incentive. You can ignore or forget every rule and still be roleplaying, possibly even having a great time. …”
Indeed, many players have had good experiences with high-authority GMs who downplay the rules and make their own judgements. I’m pretty pro-rules, and am hard to please as a player, and I’ve still had reasonable experiences under GMs using no formal rules at all.
Most multiplayer games are competitive in a direct and straightforward sense. Rpgs normally are not. Players can ignore rules because they never need to decide who’s won, nor to check that they got there by rules-legal paths.
Because published rules are often bad
Many rpg rulesets, when followed to the letter, are pretty bad, particularly those written before 2000 or so. (Rafu points out on Story Games that this can be a problem of explanation as much as actual design)
Many individual rules within otherwise good rulesets are fiddly, annoying, or just weird. Encumbrance rules are probably the most common example. A side effect of this is that players come to expect problem rules within rulesets that they have heard, or found through play, to be generally good. So they can understandably become hack-happy in this regard, discarding good rules that just seem odd at first.
A variant of the above is that many rules are not obviously valuable, and of course every rule you follow costs something (effort, head-space, blocking some interesting game event right now…). So rules that aren’t sold to players often get dropped. (Thanks to David Berg for this angle)
A common variant of this is that many rules give unrealistic or out-of-genre results. As David Berg puts it “Often one of us does know better”.
Even if a given rule is good for many groups, many rules are a bad fit for some groups. A good expression of this was by V L Darling on rpg.net —
… there are things like the actual players at the actual table, the time constraints, the set up needs. I play almost entirely online, using roll20 + discord for voice and sometimes video. I live in an asshole country with shitty internet. I have a kid and need to be out the door at a certain time. I play during the day, compared to evenings for others, so will often have to deal with phone calls or visitors, and RAW accounts for none of the real life people side of things. That’s not even accounting for more personal stuff like trauma, phobia, dislikes, whatever. All of that directly affects how RAW get applied in situations.
Because the rules as written assume they’ll be ignored
Many rulesets assume a GM who will fudge and tweak on the fly. And not all of them say that explicitly — some just assume that’s how every game is run.
There is an epidemiological effect here — it became common practice that rpgs were played like this, so it spread into most players’ model of how rpgs are played. 
Because they don’t want to change… at least not in the way the designer wants
Many players already have a way of playing that is satisfying to them, so when “trying a new game” they’re really just looking for a few new ideas that they can adopt — they have no interest in wholesale changes to how they play. (Thanks to Tommi Brander on G+ for this observation). 
Many players, even if they don’t have a satisfying way to play, are looking for one — and maybe only one. Such players are not really looking to play someone else’s game, but to build their own.
Even if they are attracted to aspects of a given rule set, players may approach a game as a tool to support their vision, rather than as a complete package. E.g. they want simple, abstract rules for space travel and world creation, but they don’t want metagame “plot point” mechanics. So if a system has all of those, they might do well to start from it and strip the latter.
(I do the above a lot. For example, I added Burning Wheel -style intent-and-stakes, and The Shadow of Yesterday -style Keys for XP progression, to the World of Darkness rules. See my post Vampire – Bristol by Night. I didn’t run Vampire RAW first, because I knew it wasn’t what I wanted.)
Because following rules is hard
Regardless of your motives and goals, following rules, learning new rules, is hard. For complex games, it’s very hard. And many players are busy adults who just want a pleasant game without having to think too hard (or are adolescents who lack the discipline to stick to anything not immediately compelling).
Understanding rules for human behaviour, especially creative human behaviour, is particularly hard, because there are so many differences between people — so many differences of assumption, so many differences of past experience. Many people have learned that trying to follow rpg rules, especially radically novel ones, is very error-prone. And thus perhaps not worth the effort.
Looking up rules during play can break the flow, so groups get used to hand-waving specific special-case-and-hard-to-remember rules.
Many people just hate rules, in any context. They’re fucked if they’re going to put up with them in a creative leisure activity.
Roleplaying game rules can be useful. Indeed, I’ve started a parallel blog post where I am collecting a list of what they do for us. But people often have good reasons to think that these rules, right now, for their group, are not going to serve them well. And in that circumstances it makes sense to fall back to “how we usually do things”.
 Chris Chinn summarises some of the relevant history on his blog. I don’t share Chinn’s obvious anger at this history, largely because I don’t believe this lead to widespread problems, at least among adult players who were otherwise emotionally stable. I don’t dispute that this lead to problems for some people. I think it caused problems for me.
 This attitude seems particularly natural for people who favour long-term play — they want continuity during something that might last for years. They might want to make small changes, and might eventually drift a long way from where they started, but they definitely don’t want the jolting shock of moving to a different game. See John Snead’s rpg.net post on this.