CWG-F278 - Using 8x8 TIP vectors when TIP size is 16x16

When the TIP block size is 16x16 is it intended that all neighboring spatial candidates and TMVP storage would use the 16x16-based vector as well?

For example, the following code doesn't seem to have any adjustments for cand_tpl_row and cand_tpl_col when the neighboring block size was 16x16.

      const int cand_tpl_row = (mi_row_cand >> TMVP_SHIFT_BITS);
      const int cand_tpl_col = (mi_col_cand >> TMVP_SHIFT_BITS);
      int_mv cand_mv = candidate->mv[0];
      get_tip_mv(cm, &cand_mv.as_mv, cand_tpl_col, cand_tpl_row, cand_tip_mvs);

So, this can end up using an 8x8 neighboring candidate vector that was not the actual one used in that corresponding neighboring block.

I didn't notice any 16x16-based adjustments for most other uses of get_tip_mv( ), so I'm wondering if all the neighboring and TMVP storage is supposed to be 8x8 based?