Why can't Postini do as well as Thunderbird junk mail controls?

[#1]

Before Postini I used only Thunderbird's junk mail controls to ditch spam messages into a spam folder. It was very effective and about 99% of the 100s/day of spam went into the junk. It learned to do this in a matter of days and I had no false positives after the first few days. The one big drawback was that the spam was downloaded onto my PC before being junked. So when Postini came along I took advantage of it so that the mail could be junked before download.
Now a lot of it does get junked by Postini before download. But Postini still lets a significant amount of spam through that then has to be junked by Thunderbird. I've been religiously notifying Postini of undetected spam using the button controls in Webmail but now it doesn't seem to get any more accurate; e.g. about 70 undetected spam messages accumulated overnight and downloaded this morning. Thunderbird junked almost all of them.
I'm grateful for Postini disposing of the bulk of the spam but I don't understand why it isn't more accurate. It seems less intelligent than Thunderbird and doesn't learn some of the spam. For instance, ever since I went onto Postini I've been getting spam with the word "adwords" in the title but Postini still lets them through no matter how often I tell it not to. Thunderbird junks them. Similarly for "Re: Hi" - I never write to people with the title "Hi" and so never get genuine mail with that in the title. Again, Postini lets them through, Thunderbird junks them.
I got Thunderbird for free. I'm sure PN pays a lot for Postini. Why isn't Postini at least as good as Thunderbird?

[#2]

How do you have your spam settings configured? You could increase the aggressiveness of the filter but you'd need to be wary of false positives if you did.
You might also want to consider enabling the Quarantine as Postini will then apply more rigorous spam checking to your account.
A bayesian filter trained to a specific customers liking (like the one Thunderbird uses) is always going to be more accurate than a service wide spam filter dealing with millions of mailboxes a day IMO.

[#3]

Bob, my Postini spam settings are the basic set because I don't want false positives. I can cope with the spam that gets through. My enquiry is not a complaint or a demand for a better service, but rather because it seems a shame that Postini is letting so much stuff through.
I'm a bit puzzled though by your comment about it being harder for a "service wide spam filter". I thought the spam training that I did was only for my account and would not affect others (and vice versa, what they set would not affect me)? If this is not so, then what I mark as spam might be marked as not spam by someone else - and surely the system wouldn't work then?

[#4]

Here's my opinion on what is or isn't spam:
Spam is unsolicited mail; mail from people or organisations that you have never passed your details to, or given anyone permission to do so.  These are generally the "increase your manhood", "buy our stock", "Hi, I'd like to meet you" type of email.  Every email you get of this type should be forwarded as an attachment to spam@spamtraining.plus.com .  (Multiple attachments are allowed, so go wild!!).
Commercial email is from organisations that you have a relationship with; Tesco, Amazon, Direct Line, Waitrose, any number of online retailers.  These are the "Get 10% off on Tuesday", "Have a discount code", "Look at our new products".  The vast majority of these companies are good citizens and provide an unsubscribe link so you can opt out of these mails if you forgot to do so when you gave them your address.  I'd suggest that these emails won't help train the spam-filter and will in fact increase the false-positive rate.
Some people/organisations fall in the crack between the two.  This is where you have to use your judgement whether or not to forward to spamtraining.
Don't forget to forward any false-positives to notspam@spamtraining.plus.com.
Hope this helps.

[#5]

But the same job is supposed to be done by using the spam/not spam buttons in webmail - or is it not?

[#6]

You are quite correct.  In webmail you've got the Spam/Not  spam buttons.
They perform the same function as forwarding the email to the addresses I mentioned.
Sorry for any confusion.

[#7]

Quote from: Bookman

But the same job is supposed to be done by using the spam/not spam buttons in webmail - or is it not?

If you are suggesting that the PN spam filters should be marking "Commercial email is from organisations that you have a relationship with; Tesco, Amazon, Direct Line, Waitrose, any number of online retailers.  These are the "Get 10% off on Tuesday", "Have a discount code", "Look at our new products"." then the answer is no!
Anybody forwarding these messages as missed spam should be shot!

[#8]

No jelv, I meant what fishter assumed - simply that the buttons should notify spam to Postini just as well as sending attachments.
But I had hoped for a response from Bob to my earlier post in this thread where I said: "I'm a bit puzzled though by your comment about it being harder for a "service wide spam filter". I thought the spam training that I did was only for my account and would not affect others (and vice versa, what they set would not affect me)? If this is not so, then what I mark as spam might be marked as not spam by someone else - and surely the system wouldn't work then?"
Why can't Postini on my account not do what Thunderbird does? I'm giving it the same training. 

[#9]

The spam/not spam training is global so the issue of what you consider is spam might not be considered spam by someone else is significant (BTW when I say global I mean GLOBAL as in all Postini users, not just the small minority who are Plusnet users on Postini) - hence my comment about shooting people!
The whitelist/blacklist in your spam settings are private to you (and incidentally can be different on each domain attached to your account).

[#10]

Now you've got me worried! How can Postini work if some people are notifying messages as spam when others want those messages? For instance, I'm a scientist and some of my work involves medical terminology which means I have to be able to receive messages with words for bits of the body that other people might consider unpleasant if they see them in an email. Some of the work is agriculture and a a word that has to occur in my emails is "rape". Most people will probably junk any emails with that word. So Postini will start junking my genuine email, won't it?
I thought Postini filters were working at the account level, not globally.

[#11]

That's why I believe the reported spam/not spam get checked before they are forwarded to Postini.

[#12]

Quote from: jelv

That's why I believe the reported spam/not spam get checked before they are forwarded to Postini.

That's right and only a small subset will get used by Postini (and even then it would be at their discretion). The best way for you to get immediate effects is to make use of the whitelisting and blacklisting controls at your disposal.