Sometimes spammers place a comment in the middle of a word, as shown in the example below:
VIA<!--text here-->GRA
This causes a single word, in this case VIAGRA, to be viewed as two words (VIA and GRA) by the e-mail client. Often, the comments themselves contain neutral words that spammers intentionally use to throw off statistical filters. The statistical filter would not catch this, because it cannot distinguish that there are HTML tags in the text. It would look for the words VIA and GRA when comparing the message to the antispam-table.txt file. Now, the HTML parser will extract the comments from the text, so that it can be examined by statistical filtering.
However, if you want embedded comments to be considered spam indicators regardless of the text, select this option under HTML Feature Filtering.