That’s called chunking. Basically, you don’t read every word in a sentence you just process units, or chunks, that you are familiar with, in this case “the mistake”. The first the is in a different chunk than the one right next to “mistake” and since that chunk is not restricting or altering proper resolution (based on your expectation), it gets scrapped at the end of processing the text.

Another culminating factor that makes it hard to spot the duplicate has something to do with eyesight. Essentially, our eyes have 2 modi: fixation and sacchades. Fixation is the standard modus and the optical nerve sends the stimuli to the brain uninterruptedly. Since the duplicates are at the end and the beginning of the next line, you have to move your eyes a longer distance to keep reading; you are performing a so-called sacchade. In order to prevent blurry sight and nausea stemming thereof when you move your eyes to another focal point (same principle like a blurry image from moving a camera while taking a picture), the optical nerve stops sending pictures to the brain during the movement. Upon reaching your new focal point, the brain backfills your memory of the travel time with the first picture it receives from the new position. This masking is called chronostasis because a very noticeable occurrence of this is that the time seems to stop for a brief moment when looking at a clock and the first second feels way longer than the following ones.