How to get Correct Results for Importance Sampling using?

Dear colleges, please give me advice for my task. I should get rare event estimation for some random function by means of Monte-Carlo simulation. In simplified form it is to estimate Probability, that min{ x[1], x[2] }< Т, where x[1] and x[2] are exponential distributed random values with mean = 1 and T<<1. I have attempted to use Importance Sampling, i.e. to use for simulation the exponential distributed random values with mean = u instead of mean =1, with values u<<1, but final result of Probability isn't correct (although the calculated Variance is very low!). For classical task (it is to estimate Probability, that min{ x[1], x[2] }> Т, where T>>1 – private task of shortest path) results are excellent, Importance Sampling using is very effective and gets us correct result. What could we do for my task? Thanks beforehand for your advice.