You should carefully read and understand this section of the documentation for glmfit, which specifies what is expected for the second input to glmfit, when the distribution is binomial.
Quoting that section, y needs to be
"y is an n-by-1 vector indicating success or failure at each observation, or an n-by-2 matrix whose first column indicates the number of successes for each observation and second column indicates the number of trials for each observation"
It seems that you have not used one of those two conventions.
