I am also a n00b in many ways, but having worked in the past for one of the major scanner companies I thought I might have something worth contributing. Please understand that I'm on this board for my hobby so please don't try and figure out which scanner company and "out" me for my past employment. My limitations of understanding come from essentially not caring what the wider market was doing if it didn't directly impact my job. Part of my job was to help my employer have the best products available for the best prices possible in the full range from the low end flat beds up to the A1 archival monsters, and all that nonsense. So if a competitor chose to sell plusteks, I could go have a beer and relax.
Your initial 4 category grouping is a good framework for figuring out what is what. It is not an easy thing to choose. You want a product designed for a purpose that isn't sold to a mass market. That means the prices are typically higher than consumer gear. It's just the sad reality, and exactly why I'm going the DIY path for my hobby book scanner. I get some e-books from google, amazon, etc, some from my local library, but the ones I really want and can't find available anywhere are already sitting on my shelves. The best book scanners in my opinion are made for libraries because they use them a lot and they want stuff that just works well and doesn't break much, especially when letting the public drive.
Inter-Library Loan and Archival are both very different from public usage. I think the first thing to look at is which style best matches what you want to do. The public kiosk stuff has to be simple, fast, and hard to break. ILL stuff often doesn't scan in color, since they come from a world where they used to photocopy and mail stuff to each other. Archival people are... anal. And I mean really, really, really anal. I don't mean that in a derogatory way either, they just have to be or they aren't doing their job. It is their job to make sure whatever they are preserving will last for a very long time. On a scale far longer than my lifetime. They want crazy high resolution, color fidelity, etc. Lots of things that in my opinion are not relevant to this board.
So if you know what kind of scanning you want to do, looking at the products aimed at that type of user should narrow down the choices.
How much scanning do you want to do? (Nothing lasts forever.)
My personal experience with plustek is not good. There is a good reason why "they" dropped them and switched to iVina a few years back. Plustek Opticbook had a MTBF of about 30,000 scans. This info isn't manufacturer claims, this info is the service department going out all over the country replacing dead units. So far the iVina unit seems to get closer to 500k scans before the sliding mechanism gives out. The cathode version isn't that bad, but the LED lighting version is pretty much un-killable and saves you the warm-up time. The iVina edge isn't 90 degree, it's more like 110. That little 20 degree extra lets the scan sensor get very close to the edge of the page and is still pretty nice to the binding. The old MFP I have at home has about an inch of plastic all around the glass scan bed. For a quick loose item scan it's fine, sucks for a book. A "book edge" scanner has the glass all the way to the edge and theoretically it scans there too.
I got the pleasure of being part of the "death-testing" on a Fujitsu SV. We rigged it up with a little software to make it scan over and over as fast as it could. From memory, the test took about 2-3 months and it was well over 1 million scans before it locked up and died. There was some software issue with the curve or keystone correction, and every so often you'd get a really messed up image for no apparent reason. It never got released in the product line up. Other than that it was awesome for the price of ~$600. Maybe a partial DIY, if you maybe could add a flat glass platen and book leveller support system underneath you'd have a really nice book scan system.
The Canon camera based systems are limited by the shutter mechanism. Atiz and others like it will need the cameras replaced somewhere around 200k to 250k scans. That's the Canon MTBF data on the shutter mechanisms and it matches up well with what I've seen in the real world. The newer EOS M line has a solid state shutter, so that could be interesting. No moving parts, maybe it can scan far more before failure. Plus the new M3 version got bumped up from 18 to 24MP, IIRC.
I happen to have an old T2i, so that's what the hobby system I'm building uses. After reading a bit here I'm considering picking up some $50 non-DSLRs and letting them wear out instead. The T2i is an actual nice camera so it seems a shame to wear it out scanning books when the $50 camera has image quality that's just as good.
I want OCR so I can read my books on an e-ink device when traveling. (LCD's fatigue my eyes pretty quickly, and the dead tree versions get heavy fast.) The tables and figures and stuff never come out that well in epub on my nook simple, so I'm not concerned about them being super high-res. My bias is good OCR, and that means 250-300 DPI usually depending on the fonts in my books. Too much DPI actually degrades the OCR accuracy for me, (I admit I could be doing something wrong), and it process much slower.
The new Canon 5DSR gets 50MP. 8000x6000 pixels, so that makes DPI and page size math easy. You can have 1000 DPI if your page is only 8x6" (and your lens can focus close enough). I think I've seen a pair of them in a scanner at ACRL, and it was A1 size, so probably only giving about 340 DPI in the end. So after spending maybe $10k on cameras and lenses, you get 340 DPI with each page being about 17x24. And since that monster was at the research library convention, it was probably selling for $50k or above.
Purists (archivists) will also endlessly debate about line vs area sensors. Hint, area sensors are far inferior. The 5DSR has some new tricks so this may be changing. Canon has a pixel pattern of red, green, blue, green (IIRC) and of those 4 pixels you get 1 pixel of "complete data". They aren't in the same physical place, so they are further inaccurate. A line sensor can do red, green, blue, and gray scale sensors sequentially and theoretically in the same physical place. I am not an archivist. I will convert to gray scale and compress, and ideally convert to OCR format like epub, so this kind of accuracy is totally irrelevant for me. Not destructively scanning the book in the first place is "preservation" enough for me. It can happily stay on the shelf for later.
Lighting is hard.
Well, any lighting is easy. Good lighting is hard. Some pages are glossy, textured, curved when not using a platen, and that means you will get spots or lines of white light washing out the page in certain places almost every time. For public kiosk use you can't realistically have a platen. People cover them with fingerprints. So, after deciding that ILL and Archival aren't the types of scanners for you, you get left with kiosk type and this is a perfect example of the v-type being optimized for the opposite of what you want. Flatbeds aren't exempt from fingerprints either. You have to be really careful to keep that glass clean or you will potentially mess up every image taken through dirty glass. So to make lighting easy you need glass, but glass isn't trouble free either... A flatbed will normally have really even reflection-free lighting.
Flatbeds are slow. Compared to "face-up", "orbital", "v-cradle", or however you call them, a flatbed takes a factor of about 20-30x more time to finish the same project. I've used them before. Scan - rotate - scan - flip & rotate - scan - rotate - scan... It sounds easy but I was always losing track and scanning pages twice or skipping some. I very much prefer the scan - flip - scan - flip - scan... work flow. Plus the actual scan time. Face up it takes between 0-1 seconds to capture 2 pages. The flat bed has to send the scan head down the bed and back, 5-20 seconds for each page depending on size and resolution settings.
With a non-DIY budget of $2000 you have some good choices. If I were in that position I'd probably go for the Fujitsu SV. iVina bookedge would be my second choice, since I hate the awfully slow scan work flow of a flatbed, and also don't personally have a need above 300 DPI.
p.s. Regarding the poor reputation of plustek, my information is mostly from 2008-2012. They weren't just bad to customers, they were bad enough to get a major reseller to dump them. They may have cleaned up their act in the last 3 years and I have no direct info about that.