The article on this site suggests that such a large number of mutations occurred in the I1 line that it appears that the population was bottlenecked for some time. Almost every I1 is I1a, and there is a long mutation list for I1, including but possibly not limited to: L759, CTS10140, CTS10338, CTS11036, CTS11042, CTS11526, CTS11552, CTS11775, CTS11783, CTS11950, CTS1393, CTS2375, CTS2524, CTS2644, CTS2738, CTS3843, CTS4130, CTS4295, CTS5167, CTS5408, CTS5513, CTS565, CTS5705, CTS571, CTS5783, CTS5891, CTS5993, CTS6140, CTS6221, CTS6395, CTS641, CTS6629, CTS7267, CTS7949, CTS8394, CTS8716, CTS9288, F3033, L509, L574, L575, L740, L750, PAGES00123, YSC0000257, YSC0000259, YSC0000264, YSC0000299, YSC0000301, L75, L80, L118, L121/S62, L124/S64, L125/S65, L157R, L187, M253, M307/P203R, P30, M450/S109, L64, L81, L123, L186, L840, P40, S63, S66, S107, S108, S110, S111
Whereas there is only DF29 for I1a, of which it's rare that any I1 is not I1a. If there were not a few people who were not I1a, then it wouldn't be labeled I1a at all and would have been considered another SNP in the I1 mutation list.
That seems to make sense in the timeline of I1a being recent and I1 beginning so long ago.
As for the specificity of ancient samples, I too wish there were more details given. When they say I1* that doesn't give us much of a clue about where in the I1 mutation list that they "branched." Maybe the paid versions of papers would have info such as that.