Often times, the phrasing used is something like "wanting to bring the baseline level of play up to a higher tier" along with "closing the gap between the top and bottom". ie: You want to bring the people who find Rabanastre hard up to the point where they start finding it easy, at which point that becomes the new baseline difficulty (rather than the comically undertuned Kugane Castle).
The thing with the skill gap is that people don't need to improve most of the time. Hence the pushing for the difficulty curve to start going upward, instead of the current weirdness where arguably the hardest dungeon in any dungeon roulette in the game is Auran Vale at level 49, followed by The Vault at 57. Those ones are serious in that if you decide to just sleepwalk your way through, they can push back. But they're not punishingly hard, either (I guess some people would say The Vault was pre-nerf, but I liked it, and groups still have trouble in there today).The skill gap will always be there, I'll admit, and unfortunately, proposing a way to shorten that skill gap enters into territory that I've been trying to avoid - which, to me, means perfecting rotations, min-maxing, spending hours learning, which is an area I'm trying hard to avoid bringing into active discussion. Conversely, I don't mind that there are top players and bottom players - my thoughts start to lean towards how to raise those bottom players up to be at the current average level in level 70 content where being average is desired.
For the 70 dungeons, Hells Lid was a step in the right direction, IMO. It's entirely clearable with a wide range of skill levels, but it also says "messing this up will hurt". It has the right idea with the DPS check phase, although it's a fairly easy check. They could keep doing stuff like that and gradually ramp up the requirement such that you'd need to be doing something resembling a proper (but not optimal) rotation to clear it. Fractal Hard doesn't accomplish the same goal. The tank can ignore half the mechanics in there and I can easily heal through it without really breaking a sweat. While that makes me feel like a great healer, it's not sending the right message.
We don't want to chase players off by sudden, jarring spikes. That's bad design. But having a sense of achievement requires the risk of failure. Beating a target dummy doesn't mean anything. I felt that way when I beat Thordan normal the first time. It was "that's it?" Comparatively, Shinryu wrecked me a few times, so when I figured him out and got that kill, it was much, much more satisfying.
The other way you can reduce a skill gap is to make the game easier to play, although lots of players frown on that kind of simplification. They went that way with some jobs in Stormblood, making mistakes less punishing in rotations and such. That helps, but more in the sense that it brings the top players down due to reducing the impact a high level of knowledge/skill will have on performance. At the same time, you don't want to have jobs so complicated that external guides and precision timing are required to be at all competent, because large swaths of the playerbase will never have that and it's frankly bad business in a mass market game.
It's a difficult balance to get right when you're trying to cater to several different types of players at the same time. I feel that's why Ultimate was a good idea: it's content I'll never do, but it's tossing a bone to the elite and saying "we know you exist and find things easy, here's something aimed squarely at you." I also thought Rabanastre was pretty good difficulty wise for content that's expected to be done by most players. And Bardam's has some good difficulty in levelling. Really, expert is where I have the biggest problem in Stormblood because the first set was just so undertuned and with that name I'd expect more than "it's literally impossible to fail this".