Having a look at the book and doing some experimenting, the problem seems to be largely caused by the leading spaces in the paragraphs and character such as the thins space.
As an example, with the following paragraph:
Code:
<p>
– Ah oui, je… » Les paroles prononcées de l’autre côté de la porte
parvenaient au visiteur mutilées, à demi étouffées. « J’ai appelé pour
annuler. Nous avons une… un… problème… ennui… de sécurité.
</p>
There is a thin space between "«" and "J". On the device, there is a much larger space than expected between "pour" and "annuler".
When I reformatted the above to:
Code:
<p>– Ah oui, je… » Les paroles prononcées de l’autre côté de la porte parvenaient au visiteur mutilées, à demi étouffées. « J’ai appelé pour annuler. Nous avons une… un… problème… ennui… de sécurité.</p>
That extra space didn't occur.
I know that there are issues in the kepub parser for some characters. For example, spacing is added around an emdash when it probably shouldn't be. And isn't in an epub rendered using the RMSDK. I think there are similar issues with other characters such as the thin space used used. But, I don't normally see these as they are uncommon in English text.
From what I can see, the extra long spacing is where there is a line-break and then leading spaces within the paragraphs and a thin space or endash on the previous line. This appears to be a bug in the parser. It might be language specific handling that is wrong. But, removing the line breaks and the leading spaces in the text improves the rendering a lot.
Note: When doing this, I didn't do anything to add the spans into the file. I took the supplied sample and added it directly to the device. And edited another copy to wrap the text within paragraphs. Adding the spans would make a difference to this. It might be a case were wrapping the paragraph rather than a sentence within works better. Looking at a sample from Kobo to see what they do might be a good idea.