why optim() gives me the initial values when I supply analytical gradient?

Question

I wrote a function to solve a maximum likelihood optimization using optim in R:

dt<-structure(list(FARM = c(1, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2, 2, 3, 
3, 3, 3, 3