I've recently been involved in some customer support efforts, and there are customers who are just unreasonable. They'll demand to have their cake, eat it too, and even get a new one. For the nuisance that a completely made up problem caused them. A problem that wouldn't even be your fault if it had been real.
They are a vast minority, but they spoil it for everyone. They consume your time and especially your team's morale. There is only so much bullshit a support agent can take before getting fed up with it and degrading their service to subsequent customers.
Now the organization has to figure out a way to detect those customers early enough to prevent them from screwing up everything for everyone. But false positives are very expensive: get one wrong and it becomes a PR nightmare.
Furthermore, if you try to give the best possible support, you must empower your agents to act. They can now screw up and even get your company in legal trouble. Good training reduces this risk, but humans making calls means errors will be made eventually.
In the end, reducing support to the bare minimum possible appears a reasonable option for many companies: it is the easiest to implement, it reduces legal/PR risks, and it has a very measurable and consistent effect (how many people stop buying/using your service after failing to get support). If that number is low enough, it just doesn't make economical sense to try to provide good support, which is a very hard endeavor for the reasons mentioned above.