Hi Derek,
I don't know what do you mean by "was able to obtain the correlations", and I'm not completely sure either about what you mean by "used this ROI to do regression". If what you did was to get some sort of average for each subject for the regions that survived the High task, and used this as an additional confound for the Low task analysis, yes, it is possible to find different results, and regions that were originally not significant may become so. It's fine.
But if you did something different than what I understood from your description, please, try to clarify a bit what was done. Thanks.
All the best,
Anderson