Get different number of sequences in denoising-stats and feature-table sums.

I did an analysis about microbiome in cancer vs normal tissue. I performed the analysis with QIIME2 and then wit R. And I have a question in rarefaction,

First: After a denoising I got the table stats.tsv and it contains de number of non-chimeric sequences by sample.

Here an example:

sample-id input filtered percentage of input passed filter denoised merged percentage of input merged non-chimeric
NT01 210543 28876 13.72 22526 15749 7.48 9172
NT02 212269 38087 17.94 33803 28788 13.56 8286
NT03 218688 63771 29.16 58613 50601 23.14 12628

In this experiment the minimum number of non-chimeric sequences was 2319, so I performed my rarefaction curve with the plugin qiime diversity alpha-rarefaction, with this command:

qiime diversity alpha-rarefaction \
  --i-table table.qza \
  --i-phylogeny rooted-tree.qza \
  --p-max-depth 2319 \
  --m-metadata-file metadata.tsv \
  --o-visualization alpha-rarefaction.qzv

And I got my rarefaction and I was happy.

Then, I import my data to R in order to use phyloseq, so I constructed my phyloseq object with my metadata, the taxonomy that I got in taxonomy.qza, the feature table that I got in table.qza and my phylogenetic tree. But when I sum the columns of feature-table I didn't get the same number of sequences than in the denoising-stats, actually it is very different. Obviusly, when I performed the rarefaction with rarefy_even depth it's also very different.

So.. the question is: Why I get different number of sequences per sample? Shouldn't it be the same result if you sum the feature-table columns (they are the samples) and the data in the column non-chimeric in the denoising-stats table?

Thanks in advance.