The author could leave off the title attribute in those cases. If no title is defined, no figure caption is rendered.
A worthwhile question to ask is: how common is it to add a title to an image but not also want to give that image a caption? There might be cases where the author wants to provide a tooltip without a caption. But it is discouraged to depend solely on the title attribute because this can cause accessibility concerns. Touch screen devices cannot display the title as a tooltip, for example. Markdown was created in a pre-iPhone world where this was less of an issue.
You could use a post-processing rule to group images together in a single figure if they are next to one another in the document (and contain titles), then generate a combined caption based on the titles of all three. There are multiple ways of doing this, so this would be outside of the spec. One example (based on the example in this HTML Doctor article) is:
![Cheeky looking Rainbow Lorikeet](/lorikeet.jpg "Rainbow Lorikeet")
<img src="/kookaburra.jpg" alt="Kooaburra" title="Kookburra">
<img src="/pelican.jpg" alt="Pelican stood on the beach" title="Pelican">
<img src="/lorikeet.jpg" alt="Cheeky looking Rainbow Lorikeet" title="Rainbow Lorikeet">
<figcaption>From left to right, Kookburra, Pelican, and Rainbow Lorikeet.</figcaption>
If you need a custom caption, a rule like this is not going to be suitable. But for a lot of cases it removes the need to remember additional syntax. It also removes to need to provide yet another way to describe the contents of the image. Do we need a standardised way of producing the shared caption? I’m not sure.