Newbie converting html to mobi - fonts screwed

[simonkbike]

Hi,
I'm converting html to mobi (for Kindle). I have some not very complex formatting changes e.g
She wore Blue Velvet
embedded in my text, which don't get to the Mobi file. - The "She" is in silver text which shows up fine in my Kindle in other books.
I've set a debug folder and discovered that the change happens between the Structure and Processed. Here's an example:
Structure html text:
<p class="MsoNormal sgc-10"><i><span class="sgc-9">Chorus 1<br/></span></i> <span class="sgc-6">Do you hear the people sing?<br/>
Singing a song of angry men?<br/>

Processed html text:
<p class="MsoTocHeading"><i class="calibre7"><span>Chorus 1<br class="WordSection"/></span></i> <span>Do you hear the people sing?<br class="WordSection"/>

You can see that the span class sgc-6 is removed from the processed file, which puts all my text the same (removes some rather important highlighting in this case instance).

Any help in how to tune Calibre to correctly interpret this text would be appreciated. I just can't seem to find anything in the help info to point me in the right direction.
Simon.