At each point information from neighbouring data is used. As the time series considered are finite, the areas at the start and end of the data (especially at higher period bands) will not have all the data required. One solution to make computation possible, chosen here, is to pad the time series with zeros where required. However, the zero padding will impact the reliability of the results. It is standard to use a cone of influence to represent this difference in reliability of results. Pale colours represent those areas outside the cone of influence with less reliable results (as seen on Fig 3). Higher period bands require more data for computation resulting in the cone shape.
ITVGMB has received Ofcom complaints after Dr Hilary Joned branded yellow card leaflets ‘a hoax’.

Baker, M. and J. Wurgler (2006), “Investor sentiment and the cross-section of stock returns”, The Journal of Finance 61 (4), 1645–1680.
Although short term relationships are erratic and sparse, this is the period band that contains most of the negative—although usually fleeting—relationships (shown by leftward facing arrows). The relationships link online activity increases to price falls (the converse is not observed). It is not surprising that occasionally discussion is associated with price falls, as negative events (e.g. blockchain bugs, and exchange hacks) are newsworthy in the community.