The UK proportion is above the median, and one of the highest in Europe. It's also widely considered by various liberal and left groups to be too high, and not doing any good. While I agree that in many cases people are being imprisoned unnecessarily, I think those cases are far outnumbered by those people who should be imprisoned but aren't.
Let's assume that, as a minimum, serial violent offenders should be imprisoned. They're not people who "lost control in a moment of madness", they're not people who could make up the harm they've caused to their victims and to society by paying a fine or doing a bit of community work: they're people who need to be temporarily removed from society for everyone else's safety. Most of them won't be imprisoned indefinitely, but they will be there for a while (and yes, a lot more work needs to be put in to rehabilitating them so that when they are released, they don't start again).
From this Shakesville post, I got to this post which references various studies. They're US studies, but since the proportion of people who will be raped is similar in the UK, I feel reasonably confident in assuming that the proportion of people who are rapists is also similar.
The more conservative estimate, from Lisak and Miller (2002)1, is that 76 of 1882 male college students (including mature students and postgraduates, looking at the details), had committed multiple rapes (and/or attempted rapes, and/or sexual assaults), none of which they had been punished already for: about 4% of the sample, and around 2/3 of the rapists in the sample. Other research finds a figure where between 5% and 15% of men are rapists depending on methodology and sample, so Lisak and Miller is definitely at the conservative end.
So, assuming that the 2/3 ratio holds generally, and McWhorter's study (McWhorter 20092) on a different sort of sample has a similar ratio, that means that between 1 in 30 and 1 in 10 of the male population are repeat undetected rapists (who will admit to it if asked in the right terms, too). McWhorter's study found no variation in these figures for any demographic group.
These are exactly the sort of unrepentant serial violent criminals that should be imprisoned... it's just not practical to do so. Doing so would - with the most conservative estimates - require increasing the prison population to more than 1 in 60 (assuming that at least some of the people already there should remain) - and possibly as high as 1 in 20.
For the UK, this would require increasing the prison population - assuming all these serial rapists were caught in a relatively short timeframe, and imprisoned for an appropriately long time, to between 1 and 3 million - roughly the population of Wales.
This is not a reasonable proportion of the population to imprison: it would require twenty times as many prisons and staff as there currently are. There is no prison system in the world that could cope with that, and I'm not sure that any society that didn't orient itself primarily around providing imprisonment could construct one - it would, for instance, mean providing 10 to 20 times more prison cells than current NHS hospital beds, but with the greatly added security needed to protect staff and inmates from the high-risk prisoners. Of course, leaving that many serial rapists free in society means a rapidly increasing number of victims, and so is also completely unacceptable.
The fact that we - quite literally - cannot capture and imprison all serial rapists (and other sexual offenders), despite the fact that we should, means that a focus on prevention: specifically on preventing men3 from becoming serial rapists in the first place, through education, through weakening or removal of the influences that cause them to become serial rapists, and through doing a far better job of capturing them earlier.
I'm not for a moment arguing that the work done in convicting rapists at the moment is not worthwhile. It is undoubtedly necessary. It's not, in itself, however, a sustainable solution, because at some point it would become too successful to support itself. Prevention is the only sustainable solution - and genuine prevention, not the worse-than-useless "be safe" advice that's currently popular, but research by Lisak and others has shown that the majority of these serial rapists have strong anti-women attitudes and a strong lack of empathy and ability to express (most) emotions. So by the time they get that far, it's probably too late: one of the figures in the government documents below is that 70% of men responded usefully to government advice on the need for consent - I'd guess that there's no noticeable overlap between that 70% and the 5-10% of serial rapists, however.
So, any prevention needs to start very early in children's education, and be reinforced continually at all levels. This is something that has not been a major focus so far. This is the 2007 plan from the government, and a recent consultation. Given the context that the studies above provide, there seem on a quick read to be some major gaps in the prevention strategy: I'll look at that more closely later.
There are a number of opinion poll results in those documents - all of which have been mentioned before - which make the scale of the problem even clearer: it is of course not just the rapists, but their apologists, who need to be considered when thinking about eliminating this crime. The rape culture is very pervasive and supportive of rapists.
1 Repeat Rape and Multiple Offending Among Undetected Rapists by David Lisak and Paul M. Miller, published in Violence and Victims, Vol 17, No. 1, 2002. I'm relying on the abstract and Thomas Millar's summarising (which I am summarising even further now).
2 Reports of Rape Reperpetration by Newly Enlisted Male Navy Personnel by Stephanie K. McWhorter, et al., published in Violence and Victims, Vol, 24, No. 2, 2009 (McWhorter 2009) - again, abstract and Millar's summary is all I have access to
3 The research concentrates on men raping women, which is by far the most common scenario (around 90% of cases). The majority of the remainder are cases where men rape other men. When a tiny reduction in the proportion of men who rape would reduce the number of victims by more than the complete elimination of rape perpetrated by women, there seems very little point in considering it. (This differs from the reasoning behind not ignoring male victims of rape, obviously)