The target noise margin is only used at the time of negotiating the connection. It isn't taken into account continuously through the life of the connection. But all sorts of things can vary according to signal conditions, so a 12 dB noise margin at two different times could be made up in quite different ways, depending on the bitloading.
But none of this explains the current absence of noise pulses. That's not the result of differences in the mode of connection, it's because something has changed so that the interference isn't there at the moment. If the noise pulses are caused by an intermittent line fault, then that intermittent fault is currently in remission, so to speak; and if it's electrical interference, then the source of that interference isn't active at the moment. My guess (no more than that) would be an intermittent line fault, mainly because I see very similar effects on my own connection.