For particular images I having an issue with reading numbers

[Neha Mittal]

I have these images on which all text is read correctly but the numbers are returning back garbage.

>tesseract darkness_11.png stdout --psm 1
OSD: Weak margin (2.74) for 28 blob text block, but using orientation anyway: 0
Background Darkness Level
Darkness: < n

I have also tried doing this with the C++ Tesseract API where I crop the region with just the number and still get similar results.
Can someone please help point out if I am missing something?

P.S. I am using eng.traineddata from the best branch.

>tesseract --version

tesseract 4.0.0-beta.1

 leptonica-1.74

  libjpeg 8d (libjpeg-turbo 1.3.0) : libpng 1.2.50 : libtiff 4.0.3 : zlib 1.2.8

 Found AVX2

 Found AVX

 Found SSE