I’m no longer working in tumour sequencing so I’m by no means an expert. But in a nutshell, the reason is that, as indicated, homozygous SNPs aren’t informative: if your allelic fraction is 100%, meaning you only observe a single nucleotide at a given position, we don’t know whether we’re dealing with a tumour related mutation. In fact, we probably don’t because the sample likely contains healthy as well as tumour cells.
By contrast, tumour-specific variants always have an AF < 100%, in other words they exhibit some degree of heterozygosity.
In fact, in healthy tissues you’d expect to see one of three cases: your variant allele (= sequence base ≠ reference) has a frequency of either 0% or 100% (homozygous), or 50% (heterozygous), ignoring rare somatic mutations. With tumour samples, by contrast, the fraction will vary depending on the penetrance of the variant in the tumour, and on the fraction of healthy to tumour cells in the sample.