In both cases, an internal link did exist on the site with the http scheme. Scrutiny treats this link as internal (as long as it has the same domain) follows it, and then all relative links will of course have the http scheme as well.
[Update - since writing this post, new functionality has been added to Scrutiny - read about that here]
I'm thinking about three things:
1. The 'Locate' function is ideal for tracing the rogue link that shunts Scrutiny (and a real user of course) over to the http site. In the shot below we can see where that happened (ringed) and so it's easy to see the offending link url, the link text and the page it appears on. Does this useful feature need to be easier to find?
2. Does a user expect that when they start at a https:// url, that an http:// link would be considered internal (and followed) or external (and not followed) ? Should this be a preference? (Possibly not needed as it's simple to add a rule that says 'do not check urls containing http://www.mysite.com)
3. Should Scrutiny alert the user if they start at an https:// url and an http:// version is found while scanning? After all, this is at the heart of the problem described above; the users assumed that all links were https:// and it wasn't obvious why they had a number of http:// links in their results.
Any thoughts welcome; email me or use the comments below.