It is because the number of data points is not exactly the same in all the curves, "missing" points are padded as zeros.
E.g. look at your initial data. Each "curve" is separated by NaN value, right? This is what I use to I split them into multicolumn matrices. But the 1st curve has 1 less data points then others, as well as the couple of last curves (see content of xdata and ydata matrices). If column has less data points than all others, these missing cells are filled with zeros by MATLAB. If we get rid of those 2 last curves and add one extra point in the very beginning of 1st curve, it seems to work OK, so the problem is with your initial data or whatever you are doing to it in a meantime.
I added 1st line below and couple of lines at the end of the code here to delete last two curves:
%% filtering
data(2:end+1,:)=data(1:end,:); % one extra point in 1 st curve (corecting missing data)
n = size(data,1);
col=1;
j=1;
xdata=[];
ydata=[];
for i=1:n
if isnan(data(i,1))
col=col+1;
j=1;
end
xdata(j,col)=data(i,1);
ydata(j,col)=data(i,2);
j=j+1;
end
[i j]=find(ydata<-0.06)
cols_to_delete=unique(j);
xdata(:,cols_to_delete)=[];
ydata(:,cols_to_delete)=[];
xdata(:,end)=[]; ydata(:,end)=[]; % getting rid of bogus last curve
xdata(:,end)=[]; ydata(:,end)=[]; % getting rid of another bogus last curve
data=[];
data=[reshape(xdata,[numel(xdata),1]),reshape(ydata,[numel(xdata),1])];
n = size(data,1); % size has now changed
