Deleting those entries makes it harder to do a deletion review because information about recent edits (such as the size of the revisions, who the recent editors are, etc.) isn't publicly available anymore. Someone can ask a sysop for a copy of the page's text from before it was deleted, but who knows, maybe the most recent revision was a heavily truncated version, where someone cut out everything controversial in an attempt to save the article. Or maybe someone gutted the article, and the reason it got deleted was that there weren't enough reliable sources left. Knowing the page length of the recent revisions would make it easier to know which version to ask for.
Eventually, recentchanges entries will get deleted anyway, so this partly addresses the argument, "If we make that data publicly accessible, it defeats the point of deleting the article."
Another problem this change would fix is that if an article is deleted, and then undeleted, the entries that were deleted from recentchanges aren't restored.