Built-in grammar checkers are like machine language translators: while they can detect some egregious errors, they are by no means capable of replacing experience and education.
In this case, the software detects I can has, which is never correct, and so flags it for correction to can have. It does not detect that cheeseburger is usually a count noun and would ordinarily take an article, e.g. I can have a cheeseburger.
But "ordinarily" is not the same as "never." It would be quite difficult to develop software that can detect such subtleties, because cheeseburger* may not in fact refer to a countable food. It could be an adjective for a type of meal. It could be used as a mass noun, say, if ground beef and cheese were mixed up and used as a pizza topping. It could refer to a flavor, as for a packaged snack food. It could refer to a red, yellow and brown color scheme. It could be a newly coined philosophical concept, or dance move, or architectural element, or anything else the author may have defined upstream in the document. And in any of those cases, yes, I can have cheeseburger would be completely correct.
Context makes all the difference.