This work describes efforts to improve the repeatability of cone calorimeter tests of polyethylene flexible foam, used for pipe insulation. This material presented significant random deformation during burning, such as warping and rolling, which led to drainage of the tar formed during material degradation and altered randomly the heating and burning processes. These effects severely undermined the repeatability of the tests and the accurate assessment of the reaction to fire characteristics of the material. Standard techniques to alleviate the problem, such as use of a retainer frame or a grid, were proven either inadequate to secure the specimen in place or undermined the accurate estimation of the material properties. The selected solution was to incorporate thin metallic pins in the sample holder, which along with the use of a retainer frame was able to eliminate the problems, producing results with good repeatability and small standard deviation.