Here’s a theory as to why: the experience may indeed be painful in the psychosocial context of our present society, but perhaps only in that context, or more specifically, because of that context.
That is, we have ideas of shame—that certain things are, or are not shameful—that are culturally based, and when we do things that offend our (learned) sense of shame, we feel, and remember, the associated negative emotions, without necessarily remembering their cause. We associate the negative emotions with the circumstance, instead of the long-gone prior that caused us to feel shameful in such circumstances. In some religions, you can feel shameful working on the Sabbath; in our society, you feel shameful having sex when society says you aren’t “ready” to. (I admit that that’s a bit of a stretched analogy.)
The more common reply to your argument, though, is that the children are reassigning a negative emotional weight to their memory of the experiences, after the fact, because the therapist/parent/whomever is expecting the experience to be negative. They don’t have to prompt for this verbally; they may be using completely neutral language, or simply asking “what happened?” Either way, their body language will show their emotional reaction to every word (and if a horse can do math based on our observed body language, we’re obviously not very good at concealing it.)
To demonstrate my meaning: If one of my friends punched me in the arm, I’d interpret that as playful at the time. If a stranger did it, I’d interpret it as hurtful. I literally feel more pain in the latter case, because of this expectation. Now, if, some time later that day, my friend insulted my race, or some other category to which I belong that implied that he just wasn’t my friend any more, I’d re-think that punch. I’d remember it hurting more.
Child abuse recountings are extreme versions of this. If you demonize the adult in the child’s mind, everything they do is going to take on a negative connotation. They’re going to start looking for the negative angle: a hug was really a rough squeeze; a toussle of the hair was really a hair-pulling, and so on. In this light, of course sex was a bad experience—it’s extremely physical with all sorts of pleasurable/painful connotations which can be switched around or played with to no end (for example, BDSM is simply a shared agreement on a set of altered connotations.)
We’re built to play games. Until we hit the formal operational stage (at puberty), we basically have a bunch of individual, contextual constraint solvers operating mostly independently in our minds, one for each “game” we understand how to play—these can be real games, or things like status interactions or hunting. Basically, each one is a separately-trained decision-theoretical agent.
The formal operational psychological stage signals a shift where these agents become unified under a single, more general constraint-solving mechanism. We begin to see the meta-rules that apply across all games: things like mathematical laws, logical principles, etc. This generalized solver is expensive to build, and expensive to run (minds are almost never inside it if they can help it, rather staying inside the constraint-solving modes relevant to particular games), but rewards use, as anyone here can attest.
When we are operating using this general solver, and we process an assertion that would suggest that we must restructure the general solver itself, we react in two ways:
Initially, we dread the idea. This is a shade of the same feeling you’d get if your significant other said, very much out of the blue and in very much the sort of tone associated with such things, “we need to talk.” Your brain is negatively reinforcing, all at once, all the pathways that led you here, way back as far as it remembers the causal chain proceeding. Your mind reels, thinking “oh crap, I should have studied [1 day ago], I shouldn’t have gone out partying [1 week ago], I should have asked friends to form a study group [at the beginning of the semester], I never should have come to this school in the first place… why did I choose this damn major?”
Second, we alienate ourselves from the source of the assertion. We don’t want to restructure; not only is it expensive, but our general solver was created as a product of the purified intersection of all experiments that led to success in all played games. That is to say, it is, without exception, the set of the most well-trusted algorithms and highly-useful abstractions in your brain. It’s basically read-only. So, like an animal lashing out when something tries to touch its wounds, our minds lash out to stop the assertion from pressing too hard against something that would be both expensive and fruitless to re-evaluate. We turn down the level of identification/trust we have with whoever or whatever made the assertion, until they no longer need to be taken seriously. Serious breaches can cause us to think of the speaker as having a completely alien mental process—this is what some people say of the experience of speaking with sociopathic serial killers, for example.
Of course, the mind can only implement the second “barrier” step when the assertion is associated with something that can vary on trust, like a person or a TV program. If it comes directly as evidence from the environment, only the first reaction remains, and intensifies increasingly as you internalize the idea that you may just have to sit down and throw out your mind.