Yeah, you're talking about something completely different than I am. Forensic paternity tests use the CODIS alleles, which count short tandem repeats (STRs) at 20 very specific non-coding loci, not across the whole genome. The whole-genome segment method of calculating genetic linkage is not statistically robust enough for forensic purposes yet (and probably won't be for a significantly long time, given how many states have passed laws against forensic genetic geneology investigations, which is preventing robust scientific validation of the methods -- there are a few manufacturers working on validating kinship analysis methods using other-than-CODIS STR loci, but it's going to take a long time to establish robust enough statistical methods).

My wife is a forensic DNA analyst. Many, many forensic DNA analysts have collected samples from multiple kinship groups they know (including their own: she's gotten profiles from her family out to second cousins) for validation studies when they're validating a new process or reagent kit! This is how I know that if you look at our forensic DNA profiles as a single comparator, the population statistics say that my sister and I are absolutely related, we are more likely to be siblings than any other relationship, but if you look at the actual profile, we could also be parent/child in either direction, and this is how I know that if you look at my wife and her niece's forensic DNA profiles as a single comparator, my wife could be my niece's parent (because she and her sister have one of the highest sibling index factors she's ever seen in her career).