Improved limits on a hypothetical <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" display="inline"><mml:mi>X</mml:mi><mml:mo stretchy="false">(</mml:mo><mml:mn>16.7</mml:mn><mml:mo stretchy="false">)</mml:mo></mml:math> boson and a dark photon decaying into <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" display="inline"><mml:msup><mml:mi>e</mml:mi><mml:mo>+</mml:mo></mml:msup><mml:msup><mml:mi>e</mml:mi><mml:mo>−</mml:mo></mml:msup></mml:math> pairs

The improved results on a direct search for a new $X(16.7\text{ }\text{ }\mathrm{MeV})$ boson that could explain the anomalous excess of ${e}^{+}{e}^{\ensuremath{-}}$ pairs observed in the decays of the excited $^{8}{\mathrm{Be}}^{*}$ nuclei (``Berillium or X17 anomaly'') are reported. Interestingly, new recent results in the nuclear transitions of another nucleus, $^{4}\mathrm{He}$, seems to support this anomaly spurring the need for an independent measurement. If the $X$ boson exists, it could be produced in the bremsstrahlung reaction ${e}^{\ensuremath{-}}Z\ensuremath{\rightarrow}{e}^{\ensuremath{-}}ZX$ by a high energy beam of electrons incident on the active target in the NA64 experiment at the CERN Super Proton Synchrotron and observed through its subsequent decay into ${e}^{+}{e}^{\ensuremath{-}}$ pairs. No evidence for such decays was found from the combined analysis of the data samples with total statistics corresponding to $8.4\ifmmode\times\else\texttimes\fi{}{10}^{10}$ electrons on target collected in 2017 and 2018. This allows one to set new limits on the $X\ensuremath{-}{e}^{\ensuremath{-}}$ coupling in the range $1.2\ifmmode\times\else\texttimes\fi{}{10}^{\ensuremath{-}4}\ensuremath{\lesssim}{\ensuremath{\epsilon}}_{e}\ensuremath{\lesssim}6.8\ifmmode\times\else\texttimes\fi{}{10}^{\ensuremath{-}4}$, excluding part of the parameter space favored by the X17 anomaly, and setting new bounds on the mixing strength of photons with dark photons (${A}^{\ensuremath{'}}$) with a mass $\ensuremath{\lesssim}24\text{ }\text{ }\mathrm{MeV}$. For the 2018 run, the setup was optimized to probe the region of parameter space characterized by a large coupling $\ensuremath{\epsilon}$. This allowed a significant improvement in sensitivity despite a relatively modest increase in statistics.