i am using a tesseract wrapper for c# (from charls, prerelease virsion), and it uses tesseract 4.
Somehow the picture i am passing is clean, but it's giving wrong results in case of letter 'S',
wrongly detects it as '5' or sometimes 'S5'
i am using following configurations
data used - tessdata-best
following flags-
load_system_dawg F
load_freq_dawg F
user_words_suffix user-words
user_patterns_suffix user-patterns
PSM mode - SingleWord
EngineMode - default
i can go for training the tesseract but seems my usecase is not that difficult to detect
following are the images and corresponding wrong text

wrong text detected -W22754

wrong text detected -W21351-2

wrong text detected -226S51
i guess these pictures are of fine quality using 300 dpi, still getting wrong results
please help if anyone find any flaw in cofig or something else.
thanks in advance.