With regard to throughput, yes XRPL is higher than Bitcoin network, though a similar number of txns to Ethereum, which A. has smart contracts B. hasn't yet integrated their Layer 1 scaling solutions (sharding, PoS) which apparently should increase throughput by many multiples.
Yes, all of them where run by Ripple to provide geographical decentralization. This is purely to achieve a high uptime and ofc demonstrate that the software works. The actual consensus mechanism ofc is(was) not decentral at all at that time. Similarly when satoshi started mining the first blocks it was also not decentral. Its kinda impossible to start decentral unless you agree with other parties to start at a specific time but technically everyone involved "colluded" for this so it kinda still defeats the point. Maybe there is a better way to start such a network but looking back, it doesn't really matter since in the beginning no one used/trusted it anyway and nothing of value could be moved since the tokens had no value so who cares if one party had full control. Currently neither Ripple nor any other entity has enough nodes to enforce or prevent a change. Last year the first amendment (a feature update) was accepted without the votes from ripples nodes. Decentralization is ofc open in one direction so its neither perfect nor finished but at lest it goes in the right direction unlike for example BTC where the mining power coming from china seems to increase year after year. AFAIK the XRPL dev team currently works on a feature that should increase reliability if a larger part of the nodes fail or get disconnected from the others. This would allow more decentralization without risking the network to halt if for example a geographical region is cut off from the rest of the network. Currently if too many nodes are disconnected at once the network would halt to prevent unintentional forking. Like if the network would be spited in halve it stops rather than both sides making different forward progress.
>though a similar number of txns to Ethereum
Actually it way higher because the XRPL has a build in DEX the vast majority of transactions are offers which usually aren't counter because they dont move value they just offer it, which may or may not result in an actual exchange and transactions that move value. But the comparison is kinda pointless anyway since XRPL Tx are almost free no one can really know which of the transaction are "meaningful" compared to eth/btc where ofc no one would make any Tx that isn't "meaningful" to them simply because it costs quite a lot.
There is also a research paper for scaling the XRPL even more but I'm p sure its implementation is on hold because it simply doesn't make much sense at the time with 5-10M Tx per day there is still 10k% grows possible and there is an existing off-chain solution for high throughput channels as well.
Also a very important factor is the cost to run it is cheap because there is a shared incentive to have performant software and energy efficient hardware. And the "people" who actually would want to run a node 24/7 are most likely companies who already have the hardware and the IT team to maintenance it etc. etc. The additional cost my be a few thousand dollars per year but that's not much for such a company.
BTW networks using this tech are running since 8+ years and consistently have more Tx than bitcoin [1] Its rather unlikely that not enough "people" could be found to keep it running.
[1] https://bitinfocharts.com/comparison/transactions-btc-eth-xr... (note that the yellow line only shows payment type Tx but there is also a DEX and other types of Tx like an offer create Tx (for trading) these are not shown since these do not exist on BTC. The total number of all Tx from all types would be way higher between 5 and 10 Millions per day.)