Everything you stated is absolutely true.
However, I think you're downplaying the costs of the hardware and support, overstating the benefits. The conversion of a one-time build charge to not only have a more expensive initial build through a particular party, but then adding in a continued maintenance cost. The next-best-alternative to the hardware contracts by having a GM within range to provide onsite support, which they already have on their payroll. With the comparative costs, they could afford having a few additional components on hand and configured to swap in case of a particular drive outage. Once again, you're right, this is cheaper on two fronts... and I realize the margins are thinner now than they used to be 5 years ago, but boutique computers, especially in the enterprise space, can run at 2x on the initial build, with 5-10% additional per month depending on service provider on an ongoing basis. The difference in costs could be used to reserve a few additional components for on-site support. Additionally... support isn't always supporting and helpful.... I'll just leave it at that.
Secondly, your statements also imply there's no redundancy in either the physical or logical layers, and any individual component crash results in an outage or loss. I'd be surprised if this is the case with seemingly identical architectures.
Ideally though, your assertions are 100% spot on... however, I think in this particular use case, their position is defendable.