LASSO implementation in Matlab is incorrect?

Matlab's Lasso() function seems to differ from textbook descriptions of Lasso. Lasso is "supposed" to add one variable at a time to the regression as the beta constraint is relaxed. But in matlab, multiple variables are often added at once. How exactly is matlab determining which variables to add? Why is it adding multiple variables simultaneously? Specifically, no matter how small I make the lambda increment and "zoom in" on a spot where variables are added, it always appears that 2+ variables are sometimes added to the equation simultaneously.