While I do not have a copy of Potassium Argon Dating, Principles, Techniques and Applications to Geochronology, a review of the text in Science Magazine reads:
"The book grew out of a pamphlet written in response to requests from
a number of US Geological Survey geologists who wanted a better
understanding of potassium-argon dating. The author attempted to
preserve the simplicity of the original pamphlet while making the book
more nearly complete and more useful. As they point out in their
preface, the book is not intended to be a scholarly or comprehensive
review of potassium-argon dating, but rather an introduction to the
principles, techniques, and applications of the method. Nevertheless
it succeeds in doing for potassium-argon dating what Willard Libby's
book Radiocarbon Dating did for that radioactive clock; it provides a
balanced and sufficiently comprehensive introduction to the subject
for the nonspecialist user of the data."
It appears that this book is a very introductory text and the quote you present from the creationist website could very likely be a part of an introductory explanation of how K-Ar dates are done. This book is written for someone who has collected samples, sent them to a lab, received a report with the lab's measurements, and wants to interpret what they're seeing. Sometimes the measurement is different than you expected, and the quote is probably about trying to figure out where exactly the disconnect is - is it with your sampling? The sample prep? Your stratigraphy? The measurement itself? If you pull it out of context, it can look incriminating, especially to the lay person, but it seems like a pretty basic statement: "if the data you got from the lab agrees with the data you have, then the data from the lab is probably good".
As for the claim that eight out of ten samples are discarded, I couldn't find a citation for that in your source. It's certainly not the case in any literature I've read. In most publications, scientists list all of the measurements that were made, including those that they ultimately left out of their data analysis, with detailed explanations as to why any dates were discarded - usually with fancy language like "we believe that the specimens may have been contaminated with modern organics during the chemical pretreatment stage", meaning "I'm pretty sure I dropped a hair in there." I've never read a paper that threw out more than 2 or 3 out of 40 dates, and I'm pretty sure that if you were tossing 8 out of 10 dates your paper would never get published in the first place. If you can find a published paper with such a high rejection rate, I would be interested in reading it - something really weird must have been going on.
On the topic of blind studies: for the following, please keep in mind that my personal experience with radiometric dating is with radiocarbon, not K-Ar. However, the general principles involved in radiometric dating apply across methods.
Blind studies are employed to prevent bias in the results. When you're measuring something qualitative (non-numeric) like increased nausea or headache, a blind study is useful because prior knowledge could alter a person's self-assessment of these kinds of factors. When you're measuring something quantitative (numeric), like the ratio of K to Ar, prior knowledge would only matter if you're deliberately tampering with the measurements.
You'd only need to employ a blind study in doing quantitative research if there was reason to believe that the people making the measurements were not only untrustworthy, but also incredibly adept at altering or manufacturing records of the data collected by instruments. In radiocarbon dating, each sample is measured "for ~5 minutes, and the complete target wheel of 32 targets is rotated a total of 6 times, giving a total analysis time on each target of about 30 minutes." The instrument records the number of isotopes reaching the counting device every second and creates hundreds of megabytes of data per sample. "We routinely use the strip chart feature of the NEC analysis code (abc) to scan the cycle by cycle data records to investigate anomalous results." That is, we look at every single data point that the instrument generates and we keep that data on file, which is available on request.
Fudging that much data in a convincing way would take effort and a depth of expertise that makes me sleepy just thinking about it. More than that, just think of the conspiracy that would be involved - just considering radiocarbon there's well over 50 radiocarbon dating labs in the world, and each lab employs tens of people including undergraduates who have no stake in an old earth, and would absolutely love to be the person who broke this vast conspiracy like something out of a Nancy Drew novel. No scientist I know has the kind of spare time or vindictiveness necessary to pull off that kind of thing.
That doesn't mean we never do blind dates. One of the times that it's done blind is in laboratory intercomparisons. In radiocarbon, for example, there have been several International Radiocarbon Intercomparisons. The most recent intercomparison, VIRI (Fifth International Radiocarbon Intercomparison), took place in 2010. Scientists at the University of Glasgow selected 7 samples which were sent to over 50 radiocarbon dating laboratories without information regarding the ages of the samples. These intercomparisions are done to learn about the variability in the measurements between labs, and so that these samples may be later used as standards. Standards are samples which have been measured repeatedly, and have an age that is well agreed upon. When measuring unknown samples, standards are always measured as well, so that you can be confident that the measurement is accurate. That is, if you can reproduce the value of the standard, you can be sure that the measurement made on the unknown is reliable.
Blind studies are not useful in standard radiometric studies, and in fact, knowing the approximate age of a sample is important so you can select appropriate standards. "To obtain a radiocarbon age the sample activity or the 14C/12C ratio must be compared to a standard material of known age." When you are measuring samples, you tune your instrument to obtain the best precision in the sample's range. So if your standard is of similar age to your unknowns, you will be able to get good precision for both your standards and your unknowns. If the measurement deviates widely from the expected value, you can use the measurement on the standard to be sure that the difference between the measured value and the expected value isn't a laboratory error.