I consider the bootstrap a very good technique for increasing the number of samples without altering the characteristics of the populations. This means that the bootstrap resampling must follow precise rules. For a correct use, the correct procedure must be used.
Your question does not allow to understand what is your case: e.g., are x1 and x2 independent with each other? Are they both dependent on a y variable? Why you need to have the same number of samples for both?