We first check how bad the MM method is compared to the QM method for some distances in the QM region

Here are the distances (Angstrom) as obtained with a QM and an MM method
  distance         qm         mm    err(mm)
 C(1)-C(2)      1.456      1.467      0.011
 C(1)-N(4)      1.147      1.157      0.010
 C(2)-H(5)      1.095      1.110      0.015

Can we get better results for the QM region with the hybrid engine?

Even though UFF has automatic atom typing, it still matters (in principle) whether we specify it on input or not
  * Without typing for each region the types are automatically guessed
  * With typing the types are always as on input (for all regions)

The only difference is in the C type for the MM region.

Here are the distances (Angstrom) as obtained with a QM and an Hybrid method without explicit typing
  distance         qm     hybrid err(hybrid)
 C(1)-C(2)      1.456      1.456      0.000
 C(1)-N(4)      1.147      1.147      0.000
 C(2)-H(5)      1.095      1.092     -0.003

Here are the distances (Angstrom) as obtained with a QM and an Hybrid method with explicit typing
  distance         qm     hybrid err(hybrid)
 C(1)-C(2)      1.456      1.456      0.000
 C(1)-N(4)      1.147      1.147      0.000
 C(2)-H(5)      1.095      1.092     -0.003

Here are some observations for this example
     * The hybrid engine does better than pure MM
     * The subtle issue whether or not we specify the types has negligible effect.