It's not literally a low pass filter with a sinc function of course, that was an analogy. It's more like a gaussian smoothing. You're not going to get ringing in the temperature record...
Here's the direct quote again if you missed it above:
"We showed that no temperature variability is preserved in our reconstruction at cycles shorter than 300 years"
Assuming a resolution period of ~120 years (full cycle) or halfcycle (assuming an effective sample every ~120 years, hence a full cycle period of ~240 years), then the recent excursion can be roughly approximated as a step function. If one lowpasses a step function, you get visible ringing. In the low resolution part of the record there is no ~120 year nor ~240 year ringing at all, so the probability of a similar Heaviside step function theta in temperature having been present but filtered out is nearly zero...