Your criticism amounts to "If it doesn't completely solve the problem for everybody its no good." and that is false.
Yes some will switch to various simple password patterns t.password for twitter... f.password for facebook... or maybe fb.password... etc. That's still an improvement. Even simple patters require some effort to break.
Some fraction will use a harder patterns that aren't immediately obvious. That's an improvement. Lets say my password is "stupidgdog" for google. Maybe your automated phishing tools will try stupidfdog on facebook... but maybe not.
Some fraction will use a slightly harder pattern.
Lets say I use stupidgHdog as my google password. My new pattern is still simple. its "stupid" + "first name of domain" + "next letter in alphabet capitalized" + "dog"
With just one sample, are you really sure your automated phishing tools going to figure out that facebook is: stupidfGdog ? And twitter its stupidtUdog?
And that's still pretty lazy as passwords go.
Some smal fraction will take the hint and use much harder patterns. That will take several fished passwords for the user and probably some human eyes to figure out. This is an improvement.
Lets say my google password is: C69.7Germanium what's my facebook password?
Here... I give you twitter on this pattern too: N47.8Vanadium.
With 2 samples passwords you've got enough of a pattern to try and brute force it... letter + 3 digits + element... 26* 1000 * 118... 2.6 million passwords to try.
Very doable if its a targeted search on a particular user... but your probably not going to spend the time looking at each fished password and then write a script to do that specific search... for just one random user. Probably.
And some fraction of people will switch to using a password safe or something, and thats an improvement too.