The length of the uplift anchors depends on the required grout to rock bond stress and hole diameter and on the length required for mass stability.  The mass stability requirement is often more critical than the required bond length.  In checking mass stability, you need to make a conservative assumption about the rock with respect to its degree of fracture.  A solid mass of rock will provide more uplift resistance than an equally hard but highly fractured rock mass of the same length.  I have rarely seen a solid rock mass when designing these anchors.  Therefore, I usually look at a rock cone and then reduce the weight to account for overlapping cones.  Don't forget to account for buoyancy and apply the appropriate safety factor.

I know of no reason there would be a minimum unbonded length.

Oh, UNBONDED! Sorry.

PTI 6.4.2 says: Generally, the free stressing length should remain unbonded after stressing, except to satisfy specific structural requirements.  PTI C6.4.2 says: Unbonded free stressing lengths allow a more flexible performance of the anchor and the averaging of structure strains resulting in less load shange in individual anchors. Typically, fully bonded and partially bonded free lengths are only applicable in massive concrete structures (dams, diaphragm wall t panels, etc).

PTI 6.8 says: The free stressing length for rock anchors and soil anchors shall not be less than 4.5 m (15 ft) for strand tendons and 3.0 m (10 ft) for bar tendons.  PTI C6.8 says:  The minimum stressing length recommended is to prevent significant subsequent reductions in transfer load due to seating losses.  Shims placed below the wedge plate can compensate for these losses.  For more information, the designer should contact the anchor tendon manufacturer.

There is even more about the free stressing (unbonded) length in PTI 6.8.

To buy PTI's Recommendations for Prestressed Rock and Soil Anchors,

http://www.techstreet.com/cgi-bin/detail?product_id=1249045

My interpretation is that anchor bolts would be grouted the entire length and post-tensioned cable would have a min. unbonded length.

Anchor bolts are not tiedown anchors.  The unbonded length assures that the anchor is properly bonded deep enough to perform as intended.  What good is bonding a tiedown anchor to the first few feet of rock under a footing?  When you test the anchor, you may be testing only the upper few feet of bond.  Therefore, you don't know for sure that the anchor is bonded down deep enough to provide sufficient mass stability.  Threadbar and strand ground anchors should have unbonded lengths  Without an unbonded length, you can't pre-stress an anchor - unless you do two-stage grouting with the second stage grout being installed after the anchor has been tested and locked off.

I just finished designing a tiedown anchor job where I needed 13' of rock bond.  The plans called for 20' minimum.  My mass stability analysis said I needed 27' of rock bond to provide a SF of 2.  If I did not provide the unbonded length, then the tendon would have to elongate to resist the design load when needed.  As the tendon elongates, the upper rock bond zone could start lifting up as the anchor pulls on the rock.  Eventually, the lower portion of the tiedown would see load.  However, too much structure and rock movement may have to occur for this to happen.

My interpretation now is determine a minimum bonded length and depth of failure cone and have 1/2 (bonded zone) below failure cone and 1/2 above and to leave the remainder unbonded provided group/spacing effects are accounted for. Also, I would interpret the minimum unbonded zone to be 10 feet for a threaded rock anchor bar based on the PTI reference.

Not quite sure I followed the example, however thanks for the post.  I do follow the idea of initial localized stresses at top of rock (almost always less compentent) gradually being transferred deeper.
 

My interpretation now is to determine bonded length required and depth of failure wedge.  Have 1/2 bonded zone above failure cone and 1/2 below with remainder unbonded.

Based on the PTI reference, I would also use a min. unbonded length of 10ft for a threaded bar rock anchor.

Not quite sure I understand the example above, however follow the idea of localized stresses at top of rock (almost always less compentent) working downward.

I have never placed half of the bonded length below the tip of the cone.  Doesn't that just make a bigger cone?