OCR works by recognizing visual characters from rendered page images. Results can vary with scan quality, handwriting, unusual fonts, page layouts, shadows, skewed pages and other image problems.
That is why the workspace shows the original page preview next to the extracted text. You can compare what was detected with what actually appears on the page before using the exported result.
A page with no recognizable text is reported as having no text detected rather than being treated as a successful extraction.