A ruler would make a terrible type designer. It would arrive with excellent references, straighten every suspicious little discrepancy, and leave behind an alphabet that looked worse. You could show it the problem, but it would insist the measurements were correct.
The trouble starts with round letters. Draw a circle and a square at the same height, and the circle tends to look smaller. In lettering, an obediently sized O can therefore look undersized beside a flat-topped neighbor. Type designers compensate by letting the round letter extend a little beyond the alignment lines.[3] This is called overshoot. The O has to be taller to look the same height.
Google Fonts' illustrated definition shows the maneuver in a lowercase o and the pointed top of an A.[1] The apparent tidy line is the result of several shapes quietly crossing it. This is excellent news if you've ever been accused of coloring outside the lines, though I wouldn't try submitting the typography defense to a building inspector.
The part I like best is that there isn't even one standard amount of cheating. In Typographic Illusions, Jonathan Hoefler explains that an S and an O can need different amounts of overshoot, judged by how they look.[3] You cannot necessarily repair the alphabet with a single new rule labeled “round things: slightly bigger.”
Nor does the trouble stop at height. Hoefler's diagrams show the middle arms of E and F sitting at different levels.[3] They look like pieces you ought to be able to reuse. But the E has a lower arm and the F doesn't, so the space around each middle stroke differs. Copying the coordinate doesn't copy the visual balance.
That is an unusually compact demonstration of expertise. A novice can make the measurements agree; the practitioner knows which disagreements the result needs. The finished letters don't announce those decisions. They simply stop looking wrong.
This isn't an argument against precision. The Glyphs sketching tutorial carefully aligns nodes, equalizes stems, and then puts the arch of an n slightly above the nominal lowercase height.[2] Measurement remains useful. It just doesn't get the final vote.
I find that more interesting than the vague instruction to “trust your eye.” There is a specific effect to look for, an adjustment to make, and a context in which to judge it. The irregularity has a job.
The next time a letter pokes a little beyond a guide, I would hesitate before tidying it up. Someone may have put it there so you wouldn't notice it.