It was already discussed several pages earlier.
We agreed that
my picture shows a successful attack vector.
Your picture shows an improbable case when a merchant accepts a transaction with very low confirmation score (no matter how many squares you'll draw keeping the same topology).
PS: Yes, mthcl fixed that issue.