If I may echo OS and Mutt .. thoughts which I suggest should be paramount ..
(a) do you have ALL the details which went into the ops engineering assessment of a particular escape procedure ? Generally, NO .. in which case you ad lib at your peril .. unless you propose to redo the entire calculation set.
(b) the critical obstacle(s) which predicate the turn can be anywhere in the procedure .. please don't presume that they are second segment .. they may well be fourth segment.
(c) OEI ... NEVER pull the power back (other than to avoid exceeding engine limits) .. at commercial weights you don't have much thrust to spare anyway .. and you want to reduce it ? (The only caveat is that some jurisdictions - eg Australia - permit a quasi visual escape for smaller aircraft so there may be a restriction to maintain visual contact but this would have a consideration for subsequent terrain and climb problems. For larger aircraft, which are done for the solid IMC case, there may be a minimum cloud base to permit visual confirmation of a turn key .. and Hobart is a good example where this can be appropriate.
(d) speed (and bank - 15 deg unless otherwise noted) determine turn radius and turn radius determines if the bumpy bit is left, right ... or dead ahead (pun intended).
(e) any decently organised flight standards department WILL REQUIRE the ops eng folk to specify a set of general rules to underpin the specific escape procedures and the pilot should be able to presume that the system captures the published requirements in the way it does business ... occasional screw-ups notwithstanding ...
(f) generally we avoid scheduling third segment turns as the dynamics are just too variable and, to capture the reasonable extremes of the potential splay, the penalties usually are too great.
(g) main point is that you, as pilot, vary the calculated procedure at your potentially DIRE peril.