You could just try to avoid using the cells, and just create one big matrix (array) with NaNs and then fill in your data. However, this would be an easy workaround just in case you have just these few columns and numbers. If you have much more numbers and columns, it could be bit trickier.
See my example:
m = NaN([3 6])
m(1:3,1) = [1 1 1]
... and go on.
Hope I helped.