Are you looking for maximum or "typical" gain values? Typically when we do coverage predictions we use typical gains since, depending on how the device is held (proximity to the body ect.) the gain can vary widely. BMCO Forums publishes a document called "Link Budgets" which has quite a lot of info on mobile/wireless receivers.

Actually the ideal antenna would have 0dBi and would not be prone to orientation problems. With cruel reality of the realistic antennas, you have a randomly oriented dipole, with or without some null-fill solution.

You can calculate your link budget with 0dBi, and deal with orientation problems as with any other fading. You'll see various diversity schemes, such as MIMO, using very same approach in gain calculations.

Antennas used in the mobile phone may be patch antenna which is RF monolithic fabricated on the board. The gain of an antenna that depends on the dimension of an antenna and current distribution of an antenna.

Every realistic antenna up to λ/2 in size will perform just like every other dipole, with maximum gain of ~2 dBi or so and corrected for losses. In link balance it is not the maximum gain that counts, but the minimum which is quite deep, and the mechanisms to mitigate these minimum problems.

In urban environment you can expect multipath propagation which is a mixed blessing. It can be harnessed using MIMO and such schemes to your advantage, or it can ruin your link pretty badly.

In rural or open areas you can not count on multipath gains, and your fancy MIMO does nothing to your advantage. Only a fixed high gain antenna aligned to a base station direction will do you good.

In mobile operation where your device is not providing any mechanism of link improvement, you depend on the base station to do it for you.

Either way, link calculations for various schemes are already available and all of them account for a randomly oriented dipole of the user equipment.

better follow ETSI TR 102 377 V1.2.1 (2005-11) and ETSI TR 101 758 V2.1.1 (2000-11) as typical values