Where do the peaks in the impulse response of a firnyquist Lth band filter come from?

When using the firnyquist built-in function to design a Lth band lowpass filter, there are peaks arising at the 2nd and 2nd last sample of the impulse response (see the figure attached). The peak does not really seem to be sensible for a lowpass filter.

That is the code:

b = firnyquist(10*6, 10, 0.1);
figure
stem(b)

Can someone help me understand where this comes from or if this is a bug?