It's probably because they're not outliers. It's just that your data follow a non-normal distribution. For example, a log-normal distribution. As you remove the most extreme observations, other observations then appear "extreme".
P.S. Don't remove "outliers". You shouldn't delete data just because it doesn't look like what you thought it should.
Thanks Sal. I agree with you. After observing the raw data I found what you mentioned here. Therefore, I resmumed the analysis without deleting them. For findings differences, I will use nonparametric tests as the data is not normal. Thanks again.
Nonparametric tests are okay. However if the data are e.g. log-normally distributed, you might consider log-transforming the dependent variable. Or, preferably, use Gamma regression. It's not as scary as it sounds. Good luck.