I know it is not super ideal; but on my WES data of n=~350 individuals I am trying to observe any cryptic relatedness that could be present in the sequencing data. To do so, I first QC my WES data in PLINK with MAF > 0.01 and call rate > 98% filters, leaving ~200K high quality variants for analysis (again, not so ideal for assessing relatedness, but I think it still should give an idea more or less). Then I calculate IBD scores in PLINK with --genome option.

We already identified a first-degree relative pair with this approach (PI-HAT = 0.58, confirmed by fingerprinting [i.e. high similarity between the lengths of STR markers all over the genome, done using stock sample tubes] and clinical files), and new sequencing results indicates another pair (PI-HAT = 0.5). However; interesting thing is that, I have two samples that are somehow related to every other individual in the cohort. I have to indicate that the cohort is consisting of multiple centers all over the Europe, presumably unrelated, so this observation is indeed false-positive.

Sample 1 -> Related to 346 others with 0.25 < PI-HAT < 0.38 scores.
Sample 2 -> Related to 346 others with 0.12 < PI-HAT < 0.22 scores.

All other pairs are having PI-HAT < 0.1.

What could be the reason that we have such an observation? First thing comes to my mind is of course a possible contamination, but I wouldn't expect that since these samples were run in batches of 12, and their libraries were prepared independently as well.

TLDR: In my whole-exome sequencing data, I find one individual with cryptic relatedness to all other 346 participants; what could be the reason?