Problem description:-

When a document-wide spell-check is launched by positioning the cursor to the beginning of the document and using F7 to locate spelling errors, it is possible to instruct the process to auto-correct all instances of a mis-spelled word. However, if the spell-check encounters a different, mis-spelled word, the previous instruction seems to be voided/lost. 

Example: consider the following text:-

-=-=-=-

The first color was red.
The second color was green.
The thrid color was blue.
The fourth color was gold.

-=-=-=-
If we wished to spell-check this text, using a British-English dictionary, we would expect all instances of "color" to be corrected to "colour", *if*, we instructed the spell-check to auto-correct when it encounters the first spelling error on the first line of text. 

However, on the 3rd line of text we have a different spelling error: "thrid" instead of "third". The spell-checker will encounter this error and offer suggested corrections as we would expect. However, after resolving this different error, the spell-checker will "forget" that it needs to auto-correct all following instances of "color" to "colour" and this instruction will need to be repeated. 

As an aside, the alternate approach, "Ignore all" does seem to work as expected - once this option is taken, all future examples of an unrecognised word will be ignored. 

Just for clarification... 

The text I am testing in this example is content that has been created by OCR-scanning some old documents using a CZUR book scanner and its integrated OCR software. I make this distinction because it is of course possible to configure Write to "auto-correct" (which I have done for "color"). In that case, auto-correct works perfectly if I type in the word incorrectly at the cursor, just not when parsing a document that contains instances of spelling errors.