I see two possible explanations:

-Most likely, the imputation changed the MAF of the variant. If this happened, the imputation may have added more observations that would have diluted the effect in that variant. Look at your MAF.

-Least likely, your P-value was multiple testing adjusted. The addition of the imputed data can have an impact as the multiple testing burden increases and thus the significant threshold becomes more stringent. Check exactly what type of significance threshold you are using.

Am I missing something?

I agree with isan,please,test SNPs window effect it may help you in explaining variance by SNPs more effectively.

I hope this help.