With your test data, I think crit damage changed from x1.5 to x2 is most possible assumption.
Say in more accuracy:
original crit damage = skill% x (phy_atkx1.5 + crit_atk) x def_factor
GT attribute crit damage = skill% x (phy_atkx2 + crit_atk) x def_factor
If so, the damage increased will be
(phy_atkx0.5)/(phy_atkx1.5+crit_atk).
If we use your test data.
(1420x0.5)/(1420x1.5+52) = 0.325
162489x(1+0.325) = 215361, which is very close to 215313 from your test.
Take your previous test for verification.
low phy atk:
(7524x0.5)/(7524x1.5+1205) = 0.301
188kx(1+0.301) = 244k
4k diff may be variance introduced by min-max phy_attack
Is this make sense?
Maybe we can try your second test but increase the crit atk to 1k~2k to see if this attribute count crit atk or not.