For 802.15.4 the range difference will not be substantially different between 2.4GHz (used worldwide) and the 902-928 MHz band (North America) or the 868 band (most of the rest of the world).

The actual propagation difference is on the order of 1 to 2%, which for your applicatoin means 1 to 2 meters out of 100 meters - really insigificant given the other variables.

At 2.4 GHz, your antenna will be much smaller for a given gain, which means you can use a larger antenna over 868/900 MHz and get better azimuth gain where it usually matters.

2.4GHz 802.15.4 has a much better selection of available chips which also means there are more components (filters, rf amps, baluns antennas, kits, applications, etc) available.

2.4GHz is more crowded with 802.11 and other wiresless applications, but generally 802.15.4/zigbee applications are not immediately co-located with these other devices. Regardless, the digital modulation techniques reduce the interference issues, and a lot of the app layer software checks for usuable channels.

Thanks,

Could you maybe explain why the propagation would basically be the same when using 802.15.4?

It's a very simple design.
Make a monocone. One cone with 45 degree angle, height of 4". Connect it to the center conductor of any connector or cable.
Here's a link of a monocone with a ground plane. Change the 22mm length to 4 inches and skip the ground plane, or add a ground plane if you want antenna patterns that tilt upward.
http://www.ensta-paristech.fr/~roblin/papers/Paper1049.pdf

Notch/vivaldi antenna is equally viable. http://en.wikipedia.org/wiki/Vivaldi-antenna