fluency illusion, judgement of learning, calibration
The title question has a direct answer: you felt ready because you tracked how smooth the material felt, not whether you could reproduce it from memory. Smoothness is a property of the page, not of your learning. The two come apart, and the gap between them is what shows up in the exam hall.
The impression that re-reading creates has a technical name, the judgement of learning: a moment-by-moment rating your brain produces about whether the material is stored. The mechanism is crude. Rather than consult long-term memory, the brain samples whatever signals are at hand right now. The most accessible signal is fluency, meaning how easily the words come off the page, how familiar a passage looks, whether the examples make sense without effort. Re-reading is essentially a fluency generator. Each pass makes the next smoother, which makes the next judgement of learning higher, with no new memory laid down.
This would be tolerable if ability scaled with confidence. It often doesn't. When you lack the knowledge, you also lack the skill to notice what you lack. The cues a skilled learner uses to spot gaps, things like "does this explanation actually answer the question" and "could I reproduce this argument from scratch", are themselves instances of the knowledge you don't have. So the lowest-comprehension readers are routinely the most confident, and the deepest readers hedge. The pattern is the opposite of what introspection suggests.
Picture twenty unfamiliar terms paired with their definitions. One evening you re-read the list for ten minutes and mark every item "known." Another evening you read it once, cover it, and write what you can from memory, peeking only after each failed try.
The two evenings feel opposite. The first ends in easy confidence; the second in frustration, with gaps visible on a mostly blank page. Two days later both groups sit a delayed test. The re-read group produces far fewer correct answers. The retrieval group, despite the rougher evening, produces substantially more correct items on average and retains a meaningful fraction of them a week later. The re-read advantage in moment-of-study confidence produces essentially no advantage on the delayed test. That is the asymmetry the dialogue was pointing at: study-time confidence is roughly uncorrelated with what you can produce days later.
The "smoothness lies" principle applies cleanly to declarative learning: facts, vocabulary, formulas, dates, named effects. It pulls less weight where genuine expertise changes the fluency signal. A surgeon who has done a procedure a thousand times can watch a routine case and judge it correctly; that fluency reflects real skill, not just exposure. Likewise in tasks that are recognition-based, such as matching a face or spotting a familiar chord progression, familiarity is the answer, and a high judgement of learning is approximately correct. The trouble is for material you are still learning. There, treat ease as a warning. Judge by what you can produce from an empty page.