Inaccurate output from Windows Screen Shot PNG OCR

[Testing Windows Screenshots]

I'm using TessBaseAPI to scan a png of WIndows screenshot,

Issue 1) List of numbers are inaccurate

7254134516423432  1324152643132424   324326178176892

text out=> 72541345 16423432  1324152643132424   324326178176892

Tesseract put a space between 5 and 1

2) Missing numbers

If there are 2 notepad windows in screenshot it fails to get the 2nd list, just fragments

Input File is a png high quality. I increase resolution to 300%.

I tried setting textord_space_size_is_variable to 1

and

I tried setting tosp_min_sane_kn_sp to different values but no difference,

How can I get the best results?

Thanks

[Testing Windows Screenshots]

Had a breakthrough - John's reply: "Gimp is your friend."

https://stackoverflow.com/questions/9480013/image-processing-to-improve-tesseract-ocr-accuracy

Note:

Gimp Scale Image: Quality: Interpolation must be set to Linear.  Cubic gave only 60% text. Now results are 95%. bottom row of numbers is appearing.

But still space between 5 and 1.