how sikuli uses tesseract

with version 1.1.0+ the tessdata folder after an approriate setup and a run, that uses the text feature, can be found at <app-data-folder>/SikulixTesseract/tessdata (see docs)

You might simply then replace the bundled standard files with your custom files keeping the standard naming.
As long this folder exists, it will not be replaced by the bundled stuff.
If it does not exists, it will be created as needed with the bundled standard stuff.

Thanks for the quick answer.
We try what you suggest.
In particular we removed all the file in <app-data-folder>/SikulixTesseract/tessdata and copied just our eng.traineddata.
The result is very different with the tessaract command specified in the question.
Any other suggestion?
Thanks in advance again.
Alfredo & Andrea

ok, then this might be caused by the very basic implementation and the fact, that it is not really adapted to the 3.0+ features (no changes the last 5 years).

I am sorry, but you have to take the feature set as is. Only version 2 later this year will have some enhancements and more options according to the current state of version 3

You might try with the Tess4J package (I will use it in version 2) or if timing is not a problem use the Tesseract command in a subprocess with files.