Prehistoric people were still evolving from a more generalized, archaic human morphology, so there's no reason to expect that they be exactly the same as their modern descendants. It's clear enough from the passage following the above quote that Stringer and McKie are not implying Negroid affinities:
It is a confused picture and suggests that racial differences were still developing relatively recently, and should be viewed as a very new part of the human condition. It is an important point, for it shows that humanity's modern African origin does not imply derivation from people like current Africans, because these populations must have also changed through the impact of evolution over the past 100,000 years.
And despite what Brace et al. conclude, their data still groups Cro-Magnon and Upper Paleolithic Europeans (blue) much closer to later and modern Western Eurasians than to Sub-Saharan Africans (red), while they acknowledge that prehistoric populations are distinguished by being "noticeably more robust than more recent human groups":
This robustness that links all prehistoric humans is likely what accounts for most misclassifications and the opinion that there lacks continuity with modern populations. But another factor is the condition of the skulls being analyzed. Jantz and Owsley (2003) found that poorly preserved, highly incomplete Upper Paleolithic crania are often misclassified as African, while those that have a large number of measurements show the expected European affinities:
Some of the discordance Van Vark et al. see between genetic and morphometric results may be attributable to their methodological choices. It is clear that the affiliation expressed by a given skull is not independent of the number of measurements taken from it. From their Table 3, it is evident that those skulls expressing Norse affinity are the most complete and have the highest number of measurements (x̄ = 50.8), while those expressing affinity to African populations (Bushman or Zulu) are the most incomplete, averaging just 16.8 measurements per skull. Use of highly incomplete or reconstructed crania may not yield a good estimate of their morphometric affinities. When one considers only those crania with 40 or more measurements, a majority express European affinity.
To examine this idea further, we use the eight Upper Paleolithic crania available from the test series of Howells (1995), all of which are complete. Our analysis of these eight, based on 55 measurements, is presented in Table 1. Using raw measurements, 6 of 8 express an affinity to Norse, and with the shape variables of Darroch and Mosimann (1985), 5 of 8 express a similarity to Norse. Using shape variables reduces the Mahalanobis distance, substantially in some cases. Typicality probabilities, particularly for the shape variables, show the crania to be fairly typical of recent populations. The results presented in Table 1 are consistent with the idea that Upper Paleolithic crania are, for the most part, larger and more generalized versions of recent Europeans. Howells (1995) reached a similar conclusion with respect to European Mesolithic crania.
That seems to be the general consensus, and Howells (1997) just comes right out and says it without mincing any words:
If Upper Paleolithic people were "European" from about 35,000 B.P., then such population distinctions are at least that old. And the Cro-Magnons were already racially European, i.e., Caucasoid. This has always been accepted because of the general appearance of the skulls: straight faces, narrow noses, and so forth. It is also possible to test this arithmetically. [...] Except for Predmosti 4, which is distant from every present and past population, all of these skulls show themselves to be closer to "Europeans" than to other peoples — Mladec and Abri Pataud comfortably so, the other two much more remotely.