I don’t see how point 2 follows, given the context. Anywhere from 30-95% of the population was dead from disease and warfare - literacy becomes a luxury in those situations.

Widespread literacy could mean a lot of things as well. Is it 2% and reserved for priests or 70% and for everyone?

Well, also, 120 years after the conquest, 100% of everyone who was alive before the conquest was dead. Transmission of complex systems of knowledge between generations generally requires organized training programs, which were labeled idolatrous, and the penalty for failing to report them to the village priest was to burn in Hell for eternity. Nevertheless, we have khipu that were made centuries later, and the article claims that in some remote areas the art did not die out until the 20th century.

I agree that knotting a khipu looks significantly slower than writing with reeds on clay tablets or with pens on papyrus, vellum, or paper. You can knit or crochet at around 5 Hz, but nålbinding, hand-sewing, and macramé are significantly slower, and the khipu looks like it would be more similar to those techniques. In https://www.youtube.com/watch?v=QFeYJ2uukrQ you can see Karin Byom making 7 nålbinding stitches in 30 seconds, about 250mHz, 20× slower than knitting or crochet. At that speed you could write about four digits per minute (250 millibits per second), compared to about 60 (4 bits per second) with west Arabic numerals and a pencil.

But there are Chinese and cuneiform characters with dozens of strokes; medieval scriptoria wrote fairly slowly, both because blackletter is fairly elaborate (especially Fraktur) and in order not to waste precious vellum; and a fair bit of the Egyptian hieroglyph corpus is literally carved into stone.

So, while slow writing systems are at a real disadvantage when it comes to producing large corpora, it's not clear that that factor alone dooms you to having a total surviving corpus that you could lift in one hand. And it certainly doesn't argue against widespread literacy, since a khipu once knotted can be read any number of times.

To be quantitative about it, the simplest wordwise probability model of English (assuming the probability that each word is independent of previous words) is about 10 bits per word, so even using the known inefficient numeric khipu encoding, you could write about 20-30 English words per hour in khipu just by assigning a number to each one. You could write down the 271 words of the Gettysburg Address in a day. This is not useful for a shopping list, but certainly for recording historical events, propaganda, contracts, proverbs, laws, or hymns.

Some of the postcolonial khipu epistles do encode spoken language, possibly inspired by Spanish writing with Latin letters, but I don't know anything about the encoding.

> Widespread literacy could mean a lot of things as well. Is it 2% and reserved for priests

Sure, if you completely redefine "widespread" to mean "very highly limited", your statement is true.

> from 30-95% of the population was dead from disease and warfare - literacy becomes a luxury in those situations.

Such a wide range means they have no idea.