I'm not sure what you mean by "I have had to set the number of samples to 700 to get the regression to run". Perhaps you are referring to the number of bootstrap samples rather than the regression itself. If you have variables that are constant within a bootstrap sample, which might well happen if, for example, you have dichotomous variables where one value is rare, so samples could be constant, this might happen. You might get around that by specifying strata variables or removing some of those rare-case variables.