The best answer would be to fix the foot switches and repeat the test so that you get accurate heel contact information if possible. It might be, if your subject is in equinus and the heel never touches the floor, that the "toe" contact is actually the start of the "gait cycle" for the subject in which case you can use the toe contact as if it were the heel.

If you are confident that heel contact did actually occur, but the heel switch failed to record it, then you could estimate the heel contact timing based on the length of stance - this is not ideal but might allow you to recover the data to some degree. Normal gait is usually symmetrical so stance timing on one side is a good estimate for timing on the other side allowing you to estimate the heel contact time. I would not recommend this approach if the subject, and reported data, has a clinical function.