Resistance does end up with the “push harder to get more power” aspect of riding outside, but it’s not what I would consider the “best simulation” either. That stems from the fact that it is essentially a linear resistance curve which is not usually what we experience outside. It happens to be more like the old magnetic dumb trainers from the past. Unless you ride at very low speeds, the nature of wind resistance as speed increases is rather different from Resistance mode.
Standard mode in TR (aka Level or Slope in other places) is a step better since it is a progressive (non-linear) resistance curve. It is more like the fluid trainers of old, because those aimed to mimic that ramped resistance level as speed increases. Kinetic Road Machine or the Lemond Revolution were the kings for feel and resistance curves of the era. New smart trainers in these modes seem set on replicating at least most of that feel.
In either case above, the reality is that once you find a gear and cadence for a steady-state effort, there is not much magic here. Assuming the rider is able to hold a cadence with a decent range (say +/-3-5 rpm) the feel of pushing that maintained power will be very similar between all trainer modes (assuming same gearing in all case). There may be very subtle deltas in the feel due to the pulsing nature of our pedal stroke, but I think this is a minor delta in the grand scheme. The larger differences seem present when you start dealing with large swings in power targets.
The demands on the rider via loading, shifting (or not) will surely vary and may be what matters to many riders. I guess it may also depend on the style of training in place as well with someone doing mostly steady-state vs someone doing more hits & recoveries may find one better than the other. We see this commonly in the on/off discussions with many pushing the non-ERG side for some good reasons, but I and others also seem happy with ERG. So the “horses for courses” can play here as much as a person’s preferences.
If “realism” is the ultimate preference here, a rider should be using a Simulation based mode via apps like Zwift, Rouvy and others since those are more “realistic” than even RES or STD. R&S end up with a static load for a given cadence and gearing as mentioned above, while Simulations here are potentially changing dynamically over time (assuming non-flat courses). These are the most like outside riding that I know, so people wanting that can use those apps and replicate the process of hitting intervals based on terrain or whatever they want.
To read some of these “this mode vs that mode” discussions, you’d expect we are talking large single-digit deltas in outcome at the very least. If that were the case, it seems to me this would be easy to demonstrate and prove which mode really is “best”. But from all I have seen in the decade plus I have used and tested these modes, there is not a clear difference in actual training impact to the rider. Plenty of anecdotes on feel but most seems to boil down to preferences. I have not seen actual science or proof of substantial differences between any of them. I’m happy to read it if anyone has related studies.
And as an aside, I also wonder how many people with the “like outside” preferences are using motion rollers, rocker plates and such? Because unless they roll outside on training wheels, there’s nothing realistic about a trainer that locks the bike in a vertical plane 
Joking there a bit, but I’d wager most people have drawn a line somewhere with their training practices and setup that falls short of a true “simulation”. I actually think that’s a GOOD thing when we do so deliberately base on consideration of the pro’s/con’s between the many choices we have. There can be very good reasons to deviate from the “outside” ideal which leads to worthwhile benefits. Most of these options have some level of compromise & benefit and I think embracing that where appropriate is better than pretending we have made no concessions despite the obvious truth that we have.
