That's a good point, and I've now corrected for that, but it makes very little difference to the top line result (that the polls underestimated the support for "no" by a large margin).
The proportion of undecided voters in these polls was about 7.1%.