7 eggs from het to het only have an about 87% chance of producing a homozygous (i.e. maybe you where just the unlucky 13%). Of course if your marker is for real you where lucky not to get any non hets out of 7 (it works out that there was the same 87% chance of at least one non het and 13% chance of all being hets or better).
IMHO the marker shows up too much in pied lines to be random. There seem to be plenty of cases of hets without the marker so obviously it isn't perfect. The big question is if there are look-alike non hets.
Of course it's a huge can of worms for those selling possible hets. I think possible hets in general are a great thing for entry-level breeders on a budget. It looks to me like reputable breeders of pieds have tried to make the sale of possible hets as fair as possible through things like selling whole clutches together or selling groups of markered and non markered babies in specific ratios.
However I'm glad to see the information about the marker discussed publicly the last few years because there has always been the potential for a shady breeder to keep the markered possible het girls back for himself of his friends and only offer the unmarkered ones to the general public as the full advertised chance possible hets. If there is anything to the marker at all the remaining non markered possible hets sold to the general public would have lower than the advertised chance of being hets. For example, out of 1000 50% possible hets about 500 will be hets. If 300 of them have the marker and it's reliable enough to raise their chances of being hets even just to 80% (I suspect it would be higher than that) then you have 240 hets in the markered group. That only leaves 260 hets in the remaining 700 unmarkered possible hets for 37% hets. In this example they are only the full 50% possible hets as a group before the markered animals are picked out. Only a random selection from the entire group is truly a 50% chance. If the marker shows up in more than 60% of the hets or is accurate at more than 80% of the time when seen (the conservative numbers I picked for this example) then the remaining odds on the non markered animals will be even worse. Sure it's a headache to deal with but we don't get to pick the way the genes actually work, only the information that we share and discuss to try to handle it as fairly as possible.