It's complicated.
Structure protection, yes of course.
Here's a beginning of the SSIF MAP of some BD. Seems that almost all right eye (dependent view=00006.m2ts, MVC stream) blocks interleaved are 0C60h*2048 bytes, and left view (AVC stream) about 2000h sectors. Does it really matter how the AVC and MVC stream relate to each other when interleaved? It would be easy to test using some statistically calculated distribution. This movie has about 2800 maps so you could just divide the file sizes by that and round it to closest one which is divident by 3*2048 (=6144, which is the size of the aligned data unit in m2ts). Might be though that if the player uses the CLPI information it wouldn't find correct 3D data.
I tested the mapping produced by MakeMKV (created the SSIF) and was able to play the movie on PowerDVD, so the UDF mapping can at least be trusted.
--------------------------------------------------
# SSIF MAP FILE v1
#
# Generated by MakeMKV v1.7.0 win(x64-release)
#
../00006.m2ts 00000000 00000C60
../00005.m2ts 00000000 00002040
../00006.m2ts 00000C60 00000C60
../00005.m2ts 00002040 00001A40
../00006.m2ts 000018C0 00000C60
../00005.m2ts 00003A80 00002040
../00006.m2ts 00002520 00000C60
../00005.m2ts 00005AC0 00001BC0
../00006.m2ts 00003180 00000C60
../00005.m2ts 00007680 00001CE0
../00006.m2ts 00003DE0 00000C60
../00005.m2ts 00009360 00001BC0
../00006.m2ts 00004A40 00000C60
../00005.m2ts 0000AF20 00002040
(and so on)