# Zero values and discontinuity in explanatory variable

One of my independent variables measures worker productivity through the variable $\frac{\log{sales}}{\text{# of workers}}$, and I'm creating one variable for skilled and another for unskilled workers. Past a point, as the number of skilled workers decreases there is a strong correlation with an increase in productivity per skilled worker, which is why I'm thinking it would create a bias if I simply let productivity be zero when the number of workers is zero. At one worker the variable has a very high value and at zero workers, it would be equal to zero. At the same time, throwing observations with zero workers of that category out is not an acceptable solution.

I'm using a GLM with an identity link function. How can I account for these observations (perhaps in a separate dummy variable?) without biasing the coefficient?