Interestingly, sometimes the same assumed "typo" is only caught once, like in this example:
Code:
<div class="images">
<img src="images/drawing1.jpg" alt="Image drawing1.jpg: Illustration: Triangles (Fig. 1 - 3)" class="calibre21"/>
<p class="image-caption">Fig. 1, 2 and 3.</p>
</div>
in which the first occurring "drawing1" is caught, but the 2nd not (at least it is not counted in the list). It looks like the preceding character might have something to do with this behavior.