It’s a stick and carrot thing.
Suppose we were to invent a punishment system from scratch. So if someone gets 20 years, they serve exactly that to the day, then they get out with 10 bucks, a new suit, some nice dental work and a couple of regrettable tattoos.
The problem with this is that with a fixed release date that is unalterable, the prisoner has absolutelly no reason to co-operate with any rehabilitation plans. Sure, if he commits a crime in prison he risks more time being added to the top, but short of that, he can just sit around becoming bitter and twisted and doing nothing by way of self-improvement. Learn to read? Learn a trade? Why bother?
So everywhere has a system of early release being available if the prisoner behaves himself, shows signs of rehabilitation, etc. The carrot. Different jurisdictions have different algorithms for calculating when earliest release will be, and different names for it, but the principle is the same, and is universal. And of course, if the prisoner doesn’t rehabilitate, he gets the stick, which is the whole sentence.
You suggest the judge should have more say in the process. Maybe. But there are a few problems. First, the judge knows a lot about the crime, but not as much as you might think about the person. There’s his criminal history, which is useful. And no doubt a bunch of psych reports. But they’re not as reliable as you’d like to think. The prisoner at the sentencing phase is highly motivated to put the best version of himself he can to the shrinks, who usually only see him for an hour or so. Psychiatrists can’t do magic.
So the people in the best position to judge release-worthiness are the people who have been watching him for years - the parole board. You can’t maintain the Boy Scout thing you did for the shrinks for very long in prison. The Board can see little details like how the prisoner reacts to provocation, the sorts of friends he makes (and declines to make), how he deals with disappointment, and so on. They wind up with a much more complete picture.
The second problem is that it is very difficult to tell what a person will be like a long time in the future. The idea that in 2009, a judge will be able to know what sort of person the community will be dealing with in 19 years is fraught with enormous difficulty, which means that fine tuning a sentence in 2009 for the release of a person in decades necessarily factors in a huge number of imponderable variables. Hence the wide range. If Prisoner 24601 works hard to deserve release his sentence should be X years. If he does not, then 2X might be a reasonable outcome for the crime committed.
You mention the problem of long sentences having a differential outcome depending on the age of the offender when sentenced. 30-50 years for a 20 year old is a helluva lot longer than it will be for Phil Spector in years actually served. That is a double edged sword, however. The person who offends late in life (or who committed an offence years ago but who is only now dealt with for the offence) gets to serve less of his life in prison than the young man. BUT (and it is a big but) he spends a much greater proportion of what is left of his life than the young man. In essence, he knows he will die in prison. So…swings and roundabouts.
The general idea is that old people sentenced to relatively short periods get a bit of a discount to reflect the fact that it is harsh to impose what is effectively a life sentence, even though it is nominally 5 years. But for those who commit offences attracting the big numbers, courts are reluctant to distort the idea of equality before the law too much to allow for that factor, or else elderly murderers would get 2 years.
Yes, the system is imperfect. But that is inevitable. We just do the best we can.