It's because $E=mc^2$.
A different way to phrase this question would be, why doesn't the binding energy of atoms compared to ions, or of molecules compared to atoms, affect their mass? And the answer is that it also does, which starts to tell you something interesting about the strong nuclear force.
If you produce some neutrons essentially at rest, and allow them to interact with protons essentially at rest, the radiation that's released as they form deuterium is about $\rm2.2\,MeV$. That's surprisingly large: the mass of the neutron and proton are each about $940\,\mathrm{MeV}/c^2$, so the energy that's released is equivalent to about 0.1% the mass of the entire system. It's right at the threshold where you could discover connection between mass and binding energy by measuring everything to three significant figures. And that's not an unusual scale. At the other end of the periodic table, when you fission uranium by adding a neutron (which works best if the neutrons are essentially at rest) the typical energy released in a fission is about $\rm200\,MeV$ --- but there are a couple hundred nucleons in a uranium nucleus, so we're still at the level of exchanging about one MeV per nucleon to reconfigure how a nucleus is arranged.
(The neutron-proton system is a counterexample to your statement that "binding energy can't be directly measured"; there the transition from the unbound state to the bound state involves emission of a single photon, whose energy is related to its wavelength in a straightforward way. The deuteron is less massive than a free proton and a free neutron; the helium-3 nucleus is less massive than a free deuteron and a free proton; and so on up the table of isotopes.)
From a chemist's perspective, imagine you have free electrons and free protons and permit them to interact. They'll also emit radiation, suggesting that each de-ionization releases about $\rm13.6\,eV$. But the hydrogen atom's mass is dominated by the proton's mass, which is still roughly $940\,\mathrm{MeV}/c^2 \approx 940\,000\,000\,\mathrm{eV}/c^2$. If you wanted to observe the effect on mass due to the electrical interaction that turns a hydrogen ion into a neutral hydrogen atom, you'd need to measure all the masses involved to ten significant figures. That's a much taller order.
And it continues to get harder if you consider molecules, where the intrinsic masses are larger and the binding energies are smaller.
The connection from here to the Higgs mechanism is surprisingly tortuous.
The Higgs is mostly responsible for the electron mass.
However, unlike nuclei compared to nucleons (and atoms compared to ions, and molecules compared to atoms), the three "valence quarks" that make up a proton or a neutron are less massive than the particle they compose. Within the nucleon, then, there are several competing things happening.
First there are an indeterminate number of quarks with finite rest mass due to the Higgs mechanism. (It's always true that the number of quarks minus the number of antiquarks is three, but the number of quark-antiquark pairs is not fixed.) Second, the interactions among these quarks force each to carry a large amount of kinetic energy --- a kinetic energy that's much larger than the energy equivalent of the rest masses of the quarks, but that's hard to quantify because, again, the number of quarks and antiquarks inside a nucleon is not well-defined. This internal kinetic energy increases the effective mass of the system.
Third, there's an additional energy barrier that must be overcome for any of the constituents of a nucleon to escape. That's the "binding energy" of a nucleon, but its definition is much murkier that the equivalent for nuclei or atoms or molecules, where the number of constituents is fixed.
You could in principle start to measure binding energies among quarks in the same way you measure binding energies among nucleons: by running reactions like
\begin{align}
\gamma + p &\leftrightarrow n + \pi^+ \\
\gamma + n &\leftrightarrow p + \pi^-
\end{align}
However, things get tricky quickly for reasons which eventually boil down to "you can't say how many quarks there are in a proton."