What leads to a divergence between the real geometry and the ideal geometry predicted by vsepr theory?

 Molecules comprised of main group elements adopt the structure that minimizes the repulsions between bond pairs (bp) of e⁻s and lone pairs (lp) of e⁻s (VSEPR theory: Wikipedia). However lp are only attracted to one +ve nucleus and consequently are more diffuse (spread out) and hence in terms of e⁻↔ e⁻ repulsion: lp-lp>lp-bp>bp-bp. So we have: CH4 AX4 regular tetrahedron angles all 109.5°; NH3 trigonal pyramidal AX3E the lone pair squeezes the H-N-H angles to <109.5° (~108°); for H2O: AX2E2 V shaped or bent - now we have lp-lp repulsion the H-O-H closes to significantly less than 109.5° (~104°). 
hope this helps

5/8 beds for each square yard