7.3.4 Solution filtering procedure
OrbitalAlternative
Very aggressive filtering was adopted in order to provide a subsample of statistically solid solutions. In particular, orbital solutions were retained which simultaneously fulfilled the following constraints:

•
fractional difference in parallax between that obtained as output of the orbital fit and that in the Gaia DR3 singlestar solution $$;

•
statistical significance of the derived semimajor axis of the orbit ${a}_{1}/{\sigma}_{{\mathrm{a}}_{1}}>20$;

•
ratio of the Gaia DR3 astrometric excess noise to the uncorrelated jitter term fitted in the orbital solution $>20$;

•
number of individual transits $>36$;

•
parallax of the source $>0.1$ mas.
The resulting selected subsample of orbital solutions has a period distribution essentially free of the doubtful values corresponding to e.g. 1 year, 6 months, the 63days precession period of the satellite, and their harmonics (see Section 7.2.5 and Figure 7.6). Overall, 629 solutions pass this selection procedure and are then successively inspected for verification and validation purposes (see Section 7.3.5).
OrbitalTargetedSearch
A primary objective of the processing run with a specific input list of sources was to evaluate the ability to retrieve robust orbital solutions in the limit of low astrometric SNR (and correspondingly low mass of the companion). The filtering procedure was therefore a partly iterative process, encompassing a combination of various filters including: a distancedependent threshold for the orbit significance, different choices for the relative agreement between the fitted parallax and the Gaia DR3 value, an upper limit to the derived value of the mass function, adaptive thresholds on the period value in units of the timespan of the observations, variable thresholds for the value of the ratio of the Gaia DR3 astrometric excess noise to the uncorrelated jitter term in the solution, and extensive model comparison with alternative, lesscomplex models to safeguard detection of bonafide candidates. In total, 783 solutions passed this initial selection procedure and were then successively inspected for verification and validation purposes (see Section 7.3.5).