How does tesseract select a character based on trust?

I am using tesseract for ocr. Tesseract has the ability to recognize text at different levels, for example, recognize single words, lines, paragraphs, symbols. There is also an opportunity to gain confidence in each of these levels. So I took a look at trust and noticed that tesseract often returns not the highest level of confidence as the best choice. Can someone please tell me how tesseract chooses the right character? for example for the character "l" it chooses neither the highest degree of confidence nor the lowest