[SPAM] ending up not marked as Junk so TB not learning what's Junk

I'm concerned I've got my Junk set up so that TB is not actually learning what is spam, plus spam is not easy to delete. First question. What is putting [SPAM] in headers before emails arrive? Is it SpamAssassin? I use my own website/email address and my host uses Spam Assassin. My host in addition has their own native Bayesian spam, but that would use *****SPAM***** headers and I never get these. The problem is this. Most junk arrives marked [SPAM] but doesn't go in the Junk folder, which is the only folder with one-click Empty Junk. So I have filters set up to direct headers with [SPAM] in into the Junk folder. (i) the [SPAM] arrives in the Junk folder not marked as Junk even though the filters are carefully set up to "Set Junk status to: Junk" (ii) it doesn't seem possible to bulk mark the contents of the Junk folder as Junk. If I click the flame icon in the header row once, then all the flame icons are switched to not-Junk as expected. But if I click it again then only some, not all, are changed back to Junk. IE completely different from normal all-off/all-on conventions. (ii) some [SPAM] ends up in the Spam folder despite the filters. This looks like pure filter failure. The goal is the get all the [SPAM] messages in the Junk folder along with the Junk the TB itself puts there, so (a) I can delete them with a single click plus (b) all marked as Junk so the TB is learning from the [SPAM] emails. How can I do that please? What am I doing wrong? Many thanks.