I tested the document itself with the ocr api, and the numbers are detected OK.
To test this further, can you please send me the small image from the area inside the pink frame? You find this image in the screenshot tab as “_lastscreenshot.png”. That is the image that is send to the OCR API for processing.
Good to know. This is the image before it is send for OCR processing.
This means the extraction of the OCR area with the pink box works correct, but the problem is with the OCR engine. => We will fix this with the next OCR updates.
Workaround: It seems the problem is because the image with the number is very small (low height, and almost no white space on the left and right side). So if you make the pink box (= area to OCR) larger, this should fix the OCR issues: