Are there matched normals for normal samples

Question

[DS]

I am reading a paper and its hypothesis is that: at least some non-reference variants in Matched Normal from tumor samples would demonstrate allele frequencys other than a Mendelian distribution of 50% or 100%.

1. But isn't matched normal limited to tumor sample? There is no point of comparing a matched normal to a normal sample?
2. non-reference variants refer to portion of sequence missing from the reference genome is present only in a small set of the population, am i correct?
3. allele frequency refers to how common an allele is in a population, so do most allele follows the Mendelian distribution of 50% or 100%?

That's many question. sorry in advance! newbie to genetic