We found that the difference was in the set-context variables where there was option to interpret end-of-line characters as :
a) CR/ LF
b) LF
We find the online tool was generating hash using LF, whereas the default for "set-context variable" was CR/LF
Once we change the set-context variables to use LF as the eol character we matched the hash256 value from the website.
Printing out the hex of the input values was useful for debugging to determine exactly what value (lf and cr and all ) is being hashed.
<Please see attached file for image>