Testosterone Didn't Rapidly Decline

"it can distinguish testosterone from any other molecule based on its exact molecular weight and fragmentation pattern. This means that does NOT fall for the bait that immunoassays like RIA do, where similar looking things are potentially tallied because antibodies are, as we call it in the business, promiscuous."

--and--

"There's no indirect measurement, no antibody binding, no relative comparison to a standard that might be off. If the machine counts 1000 testosterone molecules in your sample, that's exactly what was there."

In reality, the there are several fudge factors you need to apply to the MS to get absolute counts (starting from the back):

1. The ion detector (a pulse-counting electron multiplier) efficiency depends on the discriminator's threshold voltage, the multiplier's bias voltage, and the age of the multiplier. As the multiplier wears out from ion & electron impacts (they have a special low work-function coating on the inside of the multiplier channel), the multiplication ratio degrades. There's a precision/recall or sensitivity/specificity trade-off with respect to the bias and threshold voltages. Since these tradeoffs impact recall/sensitivity, they affect absolute counts that the detector produces for a given ion flux coming out of the mass filter.

2. The mass filter (the link shows a quadrupole mass filter, but the same applies to time-of-flight setups) may have some losses when the incoming ion fragments are not perfectly focused. This impacts absolute count rate.

3. The ionizer at the front of the MS produces ions from neutral species by hitting them with (usually) electrons (but you can use photons too). The energy of those electrons affects both the fragmentation pattern of the ions and the ionization efficiency (absolute counts, again). ("Fragmentation pattern" in this context means the mass distribution of the resulting fragments) Ionizer current also especially affects ionization efficiency (number of ions produced from a given incoming flux of neutrals). Most ionizers also have a set of ion lenses to guide the resulting fragments into the mass filter, and the potentials of these lenses are important both for focusing the ion fragments (so the mass filter has high Q) and for capture efficiency (again, absolute count rate).

Most of the things that affect absolute counts are fairly stable in a well-designed MS, but the multiplier is effectively a wear component. So you do have to compensate for that as it ages.

All this is to say:

1. Getting an absolute count from a MS requires calibration using a source with some known flux. It doesn't just come intrinsically from using a MS. I don't know enough about the older method to say if calibration is easier with a MS (almost certainly it is), but the author is hand-waving away a lot of calibration in a MS.

2. I don't know anything about the fragmentation patterns of T specifically, but I do know that organic molecules have complex fragmentation patterns, and that back-fitting to get the distribution of original species (when the incoming flux is not one pure species) can be very challenging. A MS is still sensitive (at some level) to contamination by other species.

All of this could be addressed by a self-calibration setup using a consumable (and probably only available from the original manufacturer for $$$) reference standard, but you don't get to claim that there's "no relative comparison to a standard" nor "it can distinguish testosterone from any other molecule". Both of those are false statements.