Well, you're not following the documentation precisely which says "If X is a matrix, there is one box per column; if X is a vector, there is just one box."
So, in the first case you'd have to have a label for every element in the vector and there would only be a single point for each; in the latter it did precisely what the doc says it'll do. Build the array by column instead of as a single long vector...
boxplot([Bheading.' Bheading1.' Bheading2.' Bheading3.' ],'Notch','on', ...
'Labels',{'dataset1','dataset2',dataset3',dataset4'},'Whisker',1)
While it isn't specifically the problem, it is in general poor practice to make multiple variables of the same name with subsequent indices as identifiers; use an array and indexing instead. Then such code and any other operations is/are much simpler to write and can take advantage of the power of Matlab in vector operations.
data = [Bheading.' Bheading1.' Bheading2.' Bheading3.']; % one array from multiple vectors
boxplot(data,'Notch','on','Labels',{'dataset1','dataset2',dataset3',dataset4'},'Whisker',1)
It would be best practice to create the array in the initial step(s) of the script/function rather than as shown above after having multiple names but we don't have enough information to know from whence the data came to provide specifics.
