In the series of silatranes XSi(OCH2CH2)3N, 1 (X = Me, 1a; H, 1b; F, 1c) with the known gas electron diffraction (GED) structures, the problematic geometry of 1-methylsilatrane 1a has been revised. In particular, the new value of the SiN distance (dSiN) in 1a turned out to be ∼0.06 Å longer than the generally accepted one. This dSiN resolves the long-standing contradiction between the data of the structural and spectral experiments regarding the sensitivity of 1 to the medium effect. We also performed the ab initio and DFT study of the combined series of silatranes 1a-c, silylalkylamines H3Si(CH2)3NMe2 (2a) and F3SiCH2NMe2 (2b), silylhydrazines F3SiN(Me)NMe2 (2c) and F3SiN(SiMe3)NMe2 (2d), and silyloxyamines ClH2SiONMe2 (2e,f), (F3C)F2SiONMe2 (2g,h) and F3SiONMe2 (2i), in which the GED dSiN values are in a wide range of 2-3 Å. None of the involved quantum chemical methods has succeeded in reproducing all the experimental gas-phase dSiN values in 1a-c, 2a-i with an acceptable accuracy (0.01-0.03 Å). The problems of the used methods, primarily CCSD with the Pople basis sets, are caused by four molecules with the geminal SiNN and SiON fragments (2d,f-i) and dSiN < 2.3 Å. A reasonable hierarchy of computationally accessible theory levels for studying the physicochemical manifestation of the non-covalent intramolecular SiN interactions can be constructed only at dSiN > 2.3 Å: MP2 < PBE0 ∼ B3PW91 ∼ SCS-MP2 < CCSD < CCSD(T).