Is that just something you assume is always true?
Well, I've been a networking professional for around 30 years now. I know a thing or two.
Because I haven't seen any proof of why my cellular connection would fare worse than a bandwidth-limited fiber connection.
Sorry, but there are some things that are just no longer necessary to spend time or effort proving, like fiber is better than cellular in terms of throughput and packet loss, or that the world is round. The fact you're familiar with the phrase "Can you hear me now?" should say it all. Companies that operate truly highly reliable networks don't have to advertise on the notion that your connection can remain stable long enough to make a call. If you don't want to take my word for it, there is ample technical documentation available on the internet so you can educate yourself on how the different technologies work and how the networks are operated, and then you can make your own assessment.
Another thing worth touching on is that you say you haven't experienced throttling. Well, if you start streaming bluray-type content over cellular, which is what you're talking about, you'd better prepare to experience it. Because after you stream a few 30GB movies, your cellular provider is going to throttle you, regardless of whether you have "unlimited" data or not; they will cite you for abuse of their network. People who operate these networks don't like small numbers of people soaking up vast amounts of bandwidth and causing complaints from lots of other people, so those small number of people get dealt with. They can even disable your ability to act as a hotspot entirely. They may let you buy your way out of any penalty, but they're not going to rearchitect their network according to your needs.
You only dismissed speedtest.net, so does that mean the iperf results are acceptable?
It depends what kind of iperf test you ran. Are you an expert iperf user, or did you just run the default test? iperf2 or iperf3? What command line options did you use? TCP? UDP? What target bandwidth? What packet size? What direction? The right kind of iperf test can give a better indication as to the overall characteristics of your connection, yes. At least at the moment you run the test, which is big caveat when you're talking about a shared network like your cell tower. But the best test of does a particular type of traffic work, whether it's VOIP, Videoconferencing, video streaming, or encapsulated CIFS, is to try and actually pass that type of traffic.
You did try to pass that type of traffic, and the results weren't good.
Another thing I'll mention, that perhaps isn't obvious to you, and is a caveat that both your internet connections share: It's not just about the "speed" of the local internet connection you have. In order to stream 20-40Mbps for a couple of hours at a time, that much bandwidth has to be available not just at your client-side internet connection and your server side internet connection, but at
every single hop in between constantly for the whole time. And probably on a single circuit too, because most operators will have fast switching enabled. If the server side had 10Gbps fiber in California, and the client side had 10Gbps fiber in Florida, I still wouldn't bet on being able to consistently stream 20,30,40Mbps between them for 2 hours without an interrupted stream or pixelization due to packet loss. There's a reason Netflix and Amazon don't stream at that rate.
I have a family member who streams (using Plex) to a house with a 100Mbps fiber connection, from a server hundreds of miles away that has a 1Gbps fiber connection. But there's no WiFi or Cellular involved anywhere; everything is hardwired. Some nights, they can get away with streaming blu-ray without transcoding (so that will be between 18 and 40Mbps, depending on the movie). Some nights they have to knock it down to 12Mbps. It depends on what's going on on the internet.
edit: There have been situations where I've considered transcoding, however it doesn't support seeking or embedded subtitles, so it's not really an option.
Transcoding definitely is an option, if you approach it right. And it will lead to success, which since you don't live in Seoul, is not something I would expect for streaming unconverted bluray-type content over cellular.
Video streaming natively with MC is poor, and has problems. It's a market-trailing video streamer. But all the issues you describe can be overcome in any of several ways, such as pre-converting a movie to an appropriate streaming bandwidth with Handbrake, and including the necessary subtitles, either as a substream or burned in, or by using Plex as a client/server. Or by using a different client/server altogether for streaming.
For example, I use a streaming server called Air Video HD, to stream video to my phone. It seeks better than Netflix, it does a fabulous job of adapting the bitrate to the sustained bandwidth available, and it deals with subtitles in multiple effective ways. It's so far ahead of MC in streaming it's not really comparable. Plex doesn't do quite as well, but it is free and has clients for pretty much every platform, and is still way better than MC at streaming. It seeks ok (but this depends on the platform-specific client you use; for example the AndroidTV Plex Client is absolute garbage at seeking; the Apple client seeks ok), and handles subtitles ok (at least it makes them selectable).
I'm pointing out some options because I want to help you, and for you to ultimately be satisfied and happy with the solution you arrive at. And I'm afraid your expectations for cellular just aren't reasonable. My advice is that you invest your effort in establishing a streaming workflow that functions well over the not-highly-reliable and intermittently-lossy cellular network that exists here in the US. You're obviously free to disregard my advice and go your own way.
Anyway, I hope this helps. Best of luck to you...