The relevant part of the Oxford English Dictionary entry on literal:
...[pertaining] to the etymological or the relatively primary sense of a word...".
This is the normal definition of the word in the present context (the context of classifying the various senses of a word).
So a dictionary may or may not specifically mark a certain sense of a word as literal or figurative; but that doesn't mean that it's literal when it is not specifically marked. The Oxford English Dictionary never said it would mark all non-literal senses as figurative; it only does so when whatever author who wrote or updated an entry decided that doing so might be useful to the reader. It isn't useful to mark every figurative sub-sense as figurative, if only because the large majority of senses mentioned in large entries will be figurative/transferred.
To illustrate this point, I have searched the Oxford English Dictionary for occurrences of the abbreviation lit., which often means "literal". I got around 12,000 results. An arbitrary example from the first page of results:
As you see, the first sense is marked as lit., wandering away from a physical path, as in from a forest path into the wilderness. The source of the word is Latin errare "to wander". Now, why would they mark this one as literal if all unmarked senses should be literal, as you suggested? That wouldn't make sense. So it follows that sometimes literal sense are marked as literal, while at other times they are unmarked.
Consider also how impossible it would be for sense 1 to be literal, sense 2 figurative, and then senses 3–5 (unmarked) to be literal again. You can see by the definitions and quotations of those senses that they are very far removed from the literal sense 1, as they are no longer about physically moving one's body away from a physical path; they are further removed from it than is sense 2 (figurative). They are more like further developments of sense 2, derivations.
This is the order in which the Oxford English Dictionary normally lists its senses: in rough etymological order (with side-branches). And figurative senses are always further removed etymologically than the literal sense they are based on. Hence the definition of literal I quoted above: [pertaining] to the etymological or the relatively primary sense of a word. The opposite of primary and etymological is derived.
[Additional explanation giving more background and nuance:]
Literalness is relative: the farther back you go etymologically, the more literal it gets. So strictly speaking there is only "more literal" and "less literal". As to where to stop going back, that's up to the person using the word. But one generally doesn't call a sense literal until one has gone back to the oldest sense (of the branch) that is still in use or that the dictionary states.
In arrive, the oldest sense still in use in English is that of coming physically to a place. So you could call that literal, or more literal than arrive at a conclusion. Whenever something physical (like physical movement here) disappears in a derived sense, that's a sure sign that you've arrived at a less literal sense of a word. Whenever both the physical sense and the derived (newer) non-physical sense are both still in use, it makes sense to call one literal and the other figurative when comparing them.
It's ultimately a choice; but the degree of literalness is less relative: pick fruit is no doubt more literal than pick a number. The disappearance of the physical is clear enough. So you may decide not to call pick a number literal, which may be fine in a given context; but you cannot then call pick fruit non-literal or less literal in that same context. Conversely, you can call pick fruit literal and pick a colour non-literal if it suits the context. And dictionaries are all about context: they will call a sense literal as compared to other senses mentioned, wherever it is useful to do so, usually to distinguish it from a less literal sense.
With regard to speak, it seems clear beyond doubt that its etymological and primary meaning is to utter words by emitting sound using one's mouth. It therefore seems impossible for speak to be used literally in "facts speak for themselves".