I suspect the discrepancy is largely driven by how you are defining your time use variables (TUVs) and the level of detail available in ATUS versus AHTUS. ATUS allows for much greater control over selecting with whom the respondent is spending time.
I created a TUV in AHTUS that totals time regardless of activity, time of day, or location, and requires that an “other person” is present during the activity as the “with whom” specification (but doesn’t restrict based on who else was present for the activity–the “doesn’t matter” option when creating your TUV). I compared this to three different versions of TUVs in ATUS to try to see if I could replicate the numbers. All three versions allowed for any activity, time of day, or location; I only differentiated the “with whom” options. For all three, I also left the specification for the who records I wasn’t explicitly including to “Doesn’t Matter.”
The first ATUS TUV was inclusive–literally anyone who is not in the HH could be present. The second version was very restrictive and only counts time for two types of who records: “other non-household children under 18” and “other non-household adults 18 and older”. The third version was slightly more expansive than the restrictive version; this TUV also counted time if “other-non-household family members under 18” or “other non-household family members 18 and older (including parents-in-law)” were present.
Here is the unweighted mean for my AHTUS TUV:
Here are the unweighted means for my ATUS TUVs (ordered inclusive, restrictive, in between):
As you can see, my restrictive ATUS TUV matches the AHTUS TUV I created.
Are you trying to determine if ATUS or AHTUS is more appropriate for your analysis or interested in comparing them? If you can provide more information about what you are trying to capture with the measure, I can share ideas for how to define your time use variables or leverage other features (e.g., hierarchical extracts and creating your own time use variables).