Exactly so. That was a pretty frequent vector when those images became popular, so it wasn't a crazy defensive move. Even though there are ways for criminals to defeat it, proper mitm for one, those were more complicated measures that weren't as commonly used. Higher risks, development costs, trouble with scaling, or just unnecessary, for whatever reason, the static dumb credential harvesting pages that look "legit enough" were most common.
A company with limited defensive resources could approach security like a greedy algorithm. Just constantly ask, "How are most customers at institutions similar to my own getting compromised right now? How do I prevent that with as much blunt force as possible that I can deploy as soon as possible?"
That would probably get you some bizarre defensive solutions that reduce usability. But it's not an obviously crazy general strategy.
Well, major caveat: presuming you're at least doing the basics right. If you aren't bothering with hashes, then your men are already dead.
By itself, this doesn't rule out a man-in-the-middle attack, but it might prevent an attacker from setting up bonkofamerica.com and using it capture valid username/password pairs "offline", which could be reused on the real site. Of course, this depends on people noticing (and caring) that the image is missing or incorrect, so who knows...