The general problem is the proprietary architecture and overall rent-seeking behavior. As a side effect of this, they're working against the wider ARM PC market, e.g.:

https://www.usatoday.com/story/tech/2022/11/17/arm-suing-qualcomm-keep-its-latest-technology-coming-out/10723918002/

It's so bad that they act like a cartel whenever they think they can get away with it, in seemingly bizarre ways:

https://www.accc.gov.au/media-release/arm-architecture-and-its-former-md-to-pay-penalties-for-attempted-rigging-of-university-tender

In the RISC space they are, currently, almost de facto monopolistic (we'll how RISC-V will go, hopefully). X86 is fought between AMD and Intel, this gives us a leeway on how much scummy either of the two are. ARM is very competitive because they chased many industries so far, the last one is x86: if Intel and AMD go for ARM's RISC, they will basically subdue themselves under ARM.

It's not about ARM being a good or bad company, the issue is when a company become monopolistic they are basically forced into change their founding ground.

Valve is an example of a company that tries to avoid monopolies (even self one) as much as possible.

While you have a good point about the proprietary middleware, that would be far from inherently unportable - just not yet ported, likely due to low profitability. Something truly unportable would rely on specific hardware quirks, like - and I'm very sorry I can't find the original source - one old instance of some game that relied on the timing of a spinning drumhead in order to select the correct data or instruction after the completion of one prior; and even then, a port could be made with sufficient understanding of the original system.