Is the following mechanism correct: Legacy spam transactions would been recognized and avoided because they would "steal" them computing power for no benefit?
"Steal the computing power" is a pretty weird way to label this. I'd rather say that legacy transaction spam attacks would be recognized, and pools/miners could start prioritizing the other set of transactions.
Yep, maybe. What I meant was that big spam transactions would cost them an amount of hashing power that they - if they ignored these transactions or give them a very low priority - could use better to find blocks faster and get an advantage over their competitors.