git merges work with different states of the content being tracked, and not with individual files. To paraphrase, git does not see a fundamental difference between a commit involving two files or a commit involving one file.
Therefore, there is an expectation in git workflow that there is a thematic consistency in commit and branch strategies. More specifically, it is more "gitty" to create a commit that commits a "subject" than a commit that commits a specific file. It is more "gitty" to have branches that contain work performed on specific "topics" rather than branches aimed at developing specific files.
Referring back to your question, without knowing too much about your code or git strategy, it is possible that the need to merge a particular branch in different ways depending on the state of individual files indicates that the branching model may need a little polishing.
If you need changes to the release branches, but not changes to the README.md file in the release branch, it is possible that the changes to README.md from the release may need to be split so that you can merge the release without merging the changes to README.md . However, given that the branch is called release, I assume that you do not want to hit it hard.
If you need to merge with the release, but ignore the README.md changes, this is a repeat, you might consider creating a parallel branch for the release, for example release-noreadme. You can then hack everything you want into release-noreadme, including massaging the README.md file according to your needs, so that you can merge with release-noreadme without worrying about tiptoe exchanging specific changes in certain files.
tl; dr: consider synchronizing a branch where you can smooth out unwanted changes, then merge from that branch instead of releasing.
source share