With apologies for longposting and to an extent, repeating a couple things:
100%.
I very much understand why LEM wrote what they wrote, because at the time it was about maximizing then-modern functionality per dollar of spend, and, hands down, if you're at Computer Renaissance in 1997 and a 6200/75 and 7200/75 are both sitting in front of you and each cost $500 or so, the 7200 is a better deal.
But it's not 1997 any more and it hasn't been for over 25 years and the value in the 5200/6200 from a vintage computing standpoint, I would personally argue, is that most people's tourism ends up centering around things these machines do very well - ClarisWorks, HyperCard, basic OS navigation, and period edutainment or other low end games that either run well regardless or are specifically targeted at this hardware.
As the number of working machines dwindles, if someone doesn't have specific needs, the 5200/6200 is a great option for someone who wants to touch that specific nostalgia or want something that does most basic stuff fine.
(Of course, there's also just QEMU-PPC so it can depend on which specific things you want to tour, I suppose.)
Pulling back, I'd say low end systems (this applies to all hobbies) have a genuine spot in the hobby, because not everybody who's vintage computing is doing high performance technical computing, or necessarily even, say, playing the highest-end games or whatever.
And, so I'd say it's great that systems like these are around to meet the needs of people whose needs they can meet.
So, two major points.
1. LEM is great, however they have spent over 25 years doubling down on a lot of stuff about the 5200/6200 that is just straight-up factually incorrect. Literally to this very day the 6200 page still claims it doesn't have hardware handshaking on its serial ports, which is false. They left "left32/right32" up until at least 2022 or 2023, and it's probably still on their page somewhere.
If you can distinguish between what's purchasing advice for if it's 1997 and what's useful in today's context, and weed the actual truth out of stuff they just fabricated for whatever reason, then yes the site's great.
LEM also fails to acknowledge that, say, the PB5300/2300/1400, which IIRC it doesn't categorize for this reason, is the same underlying "68k LC platform with a preintegrated PPC upgrade" architecture.
2. LEM doesn't really acknowledge this the 6200/75 and a 6300/100/120 are AFAIK fairly different experiences, for legitimate and real technical reasons. The main problem with the 75MHz 603 is that it has too little L1 cache When doing 68k emulation, especially with the stock 7.5.x releases, lots of work exceeds that cache's size and the machine spends a bunch of time waiting for stuff to enter/leave cache. All faster 603s have more.
System7Today recommends Speed Doubler 8 or prioritizing running PPC-native code to avoid that problem, and when you do, the 6200/75 benches and feels the same as a 6100/66 or so. IME on my own 6200/75, 9.1 feels way slower overall than 7.6.1, even though if you fire up MacBench 4, the 9.1 install will return faster benchmark results.
The other beat LEM missed is that the PowerBooks 5300/2300 and 1400 all use the same underlying architecture, albeit with faster CPUs that almost all have the bigger L1 cache (but with their own problems, may have less or no L2 cache) and with some parts swapped to new parts.
(But the 1400/166, 6300/120, and 5200/75 all share the common architecture of being a Mac 630 platform with a PPC603 upgrade pre-integrated.)
The main downside even once you get to machines that are much faster and don't have the L1 problem is that the PCI machines are just faster. 7600/120 outbenches a 1400/166. And of course, as you look to newer software the ability to run more RAM and use PCI hardware helps.
But - speed isn't always everything and even if they win at MacBench numbers a 7600/120, 9600/300, or a G4@2GHz doesn't do anything meaningful in ClarisWorks 4, HyperCard 2.x or Oregon Trail any faster than a 6200. My 1400 does everything I need it to do fast enough even though a 2400/3400 or even 7200 would be faster, etc etc.
Applying that standard to a "road apple" would make it totally incapacitated after about 5 years.
Some of this is just that The Plateau means computers last longer now.
But I'd fully agree that a lot of what LEM defines as a road apple maybe makes sense in 1997 but not today.
When I joined here and AppleFritter in 2002, on either my Performa 578 or my 840av (forget which TBH, but both would've been ~8 years old) both were considered resoundingly obsolete by modern standards because of office document compatibility and because v5 and v6 web browsers were becoming the norm and 68k support was left behind on v4. Plus, low memory ceilings and slow systems overall meant that multi-tasking was tougher even on a high end 68k than it might have been on a midrange PPC or something then-new.
OTOH today a moderately upgraded 10-15 year old system has no real trouble keeping up with basic tasks. The only reasons I'm upgrading from a 10 y/o system to a 7 y/o one are to stay on the Windows 11 HCL, but even that's bypassable in a few ways without much change to a given piece of hardware. (And of course: Linux.)
My personal day-to-day is easily doable on a system from ~2010, albeit with upgrades like a quad-core CPU, 8-16GB of RAM a discrete GPU and an SSD.